www.elsevier.com/locate/yjmcc

Journal of Molecular and Cellular Cardiology 40 (2006) 520–528

Original article

* CorresponE-mail ad

0022-2828/$doi:10.1016/j

Human umbilical vein endothelial cells fuse with

cardiomyocytes but do not activate cardiac gene expression

Robert E. Welikson a, Stefanie Kaestner a, Hans Reinecke b, Stephen D. Hauschka a,*

aDepartment of Biochemistry, University of Washington, Seattle, WA 98195, USAbDepartment of Pathology, University of Washington, Seattle, WA 98195, USA

Received 17 August 2005; received in revised form 21 December 2005; accepted 10 January 2006Available online 06 March 2006

Dedicated to the memory of Stefanie Kaestner, whose many contributions and meticulous experimentation are greatly missed

Abstract

It was recently reported that human umbilical endothelial vein cells (HUVECs) transdifferentiate and express cardiac genes when co-culturedwith rat neonatal cardiomyocytes (Condorelli et al. (2001)). If substantiated and optimized, this phenomenon could have many therapeutic ap-plications. We re-investigated the HUVEC-rat neonatal cardiomyocyte co-culture system but detected cardiomyocyte markers (sarcomeric myo-sin) in only 1.2% of the cells containing nuclei that were immuno-positive for human nuclear antigen (HNA); and the frequency of such cells wasnot enhanced in co-cultures containing more embryonic cardiomyocytes. Because the majority of HNA-positive/myosin-positive cells were bi-nucleated, we tested the hypothesis that these cells resulted from HUVEC-cardiomyocyte fusion rather than from HUVEC transdifferentiation.Analysis with a Cre/lox recombination assay indicated that virtually all HUVECs containing cardiac markers had fused with cardiomyocytes. Todetermine whether human cardiomyocyte genes are activated at low levels in HUVEC-cardiomyocyte co-cultures, quantitative RT-PCR wasperformed with primers specific for human β-MyHC and cTnI. We found no evidence for transcriptional activation of these genes. None ofour data support conversion of HUVECs to cardiomyocytes.© 2006 Elsevier Ltd. All rights reserved.

Keywords: Heart; Cardiomyocyte; Endothelial Cell; HUVEC; Fusion; Cre/lox; Real Time PCR

1. Introduction

Recent studies have concluded that bone marrow cells [1,2],resident endothelial cells [3], circulating endothelial cells [4,5]and a putative novel resident heart cell type [6–8] are capableof converting to cardiomyocyte-like cells within injured hearts.Condorelli and co-workers also reported that human umbilicalvein endothelial cells (HUVECs) transdifferentiate to a cardio-myocyte phenotype when co-cultured with rat cardiomyocytes.While the molecular mechanisms of transdifferentiation are un-known, if substantiated and further optimized, conversion ofendothelial cells into cardiomyocytes could have many thera-peutic implications.

Studies of chick development in our lab have shown that asmall sub-population of myosin expressing cells (~3%) in the

ding author. Tel.: 206 543 6062, fax: 206 685 1795.dress: haus@u.washington.edu (S.D. Hauschka).

- see front matter © 2006 Elsevier Ltd. All rights reserved..yjmcc.2006.01.009

E3 tubular heart also contain von Willebrand factor (vWF), anendothelial cell marker [9]. Similar observations have been re-ported for immunostained cultures of quail embryo pre-cardiaccells [10,11]. These studies suggest that environments withinthe developing heart may at least temporarily permit some pre-cursor cells to follow either cardiomyocyte or endothelial celldevelopmental programs. If so, we reasoned that cues from amixture of embryonic heart cells might be more potent withrespect to inducing HUVECs to enter a cardiomyocyte cell pro-gram.

An alternative explanation for cells that exhibit multiplephenotypes is that they arise from fusion of two different celltypes. For example, bone marrow cells exhibit low levels ofcell fusion with liver, brain and heart cells [12–15]. Putativecardiac progenitor cells [6,16] are also capable of fusing withcardiomyocytes when injected into injured myocardium [7],and skeletal muscle cells can fuse with cardiomyocytes wheninjected into mouse hearts [17]. Additionally, just prior to the

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528 521

completion of our studies Matsurra et al., 2004 published stu-dies that are consistent with the ability of cardiomyocytes tofuse with endothelial cells [18].

To test whether HUVECs can transdifferentiate to cardio-myocytes [3], or whether the dual phenotype cells result fromcell fusion, we cultured HUVECs with rat neonatal cardiomyo-cytes as well as with the mixed cell populations present in E3chick hearts. In all experimental combinations the co-culturecells were marked so that accurate discrimination between pu-tative transdifferentiated HUVECs and HUVECs that hadfused with cardiomyocytes was feasible. Although we ob-served a very low percentage of cells that were immuno-posi-tive for both endothelial and cardiomyocyte proteins, theHUVEC transdifferentiation hypothesis was not supported; alldata related to dual phenotype cells could be accounted for byHUVEC-cardiomyocyte fusion. Additionally, quantitativeRT-PCR analysis of HUVEC-cardiomyocyte co-cultures indi-cated no activation of human cardiac muscle gene expressionfrom the endothelial cell-derived nuclei.

2. Methods

2.1. Cell culture

Neonatal and embryonic heart cells. Neonatal cardiomyo-cytes were prepared from the ventricles of 2- to 3-day-oldSprague-Dawley rats as previously described (Iwaki, et al.,[19]. Rat cardiomyocytes were maintained in DMEM, contain-ing 10% fetal bovine serum (FBS). Chick cardiac cell cultureswere prepared from E3 embryos. Typically, 100 tubular heartswere collected and stored at room temperature in a 100 mmPetri dish containing 10 ml of complete media 89% DMEM,containing 10% fetal bovine serum (FBS), 1% chick embryoextract (CEE), 50 U/ml penicillin, 50 μg/ml streptomycin(pen/strep) and 250 ng/ml fungizone. Collecting hearts re-quired about 2.5 h. The hearts were then rinsed twice in phos-phate buffered saline (PBS) and incubated for 5 min in 0.5 ml0.05% trypsin at 37 °C in a 5% CO2 humidified incubator. Thecells were dispersed by triturating ~ 10 times with a P200 pi-petman and the partially dissociated tissues were then incu-bated for an additional 5 min. After the second incubation,the tissue pieces were again triturated about 10 times and thesuspension was then checked under a microscope to confirmthat the cells were dissociated. An equal volume of completeculture medium was added to the suspension and the cells werethen plated onto tissue culture plates or chamber slides, thathad been adsorbed overnight with 20 μg/ml fibronectin (Sigma,St. Louis, MO).

For experiments in which chick and rat cardiomyocyteswere to be co-cultured with human umbilical vein endothelialcells (HUVECs), the heart cells were isolated as describedabove and cultured on fibronectin coated 100 mm plates inMedium 200, containing penicillin–streptomycin (pen/strep)and a 50-fold dilution of low serum growth supplement (50X LSGS, Cascade Biologics, Portland, OR, cat. # S-003-10).The cells were treated with 1 μg/ml mitomycin C (Sigma) for

12–16 h to prevent fibroblast overgrowth. The cultures werethen washed 6 times with PBS and re-fed with 10 ml Medium200 containing LSGS and cultured for an additional 1 h priorto addition of HUVECs.

For some experiments, co-cultures with rat neonatal cardio-myocytes were performed exactly as described in Condorelli etal. [3]. In these experiments the heart cells were maintained inDMEM, containing 10% FBS for 2 days, and then switched toDMEM containing 5% horse serum. Under these conditions, intwo separate experiments, we found no HUVECs containingmyosin. Therefore we used the media conditions describedabove for our data collection and quantitative analyses.

Human umbilical vein cells. Pooled isolates of HUVECs,obtained from Cascade Biologics, Portland, OR, were ex-panded in Medium 200, containing pen/strep and LSGS. Cul-tures underwent four population doublings before they werecryopreserved in multiple aliquots. Prior to co-culture experi-ments, the HUVECs underwent 2-3 further doublings. Cellswere trypsinized, resuspended in complete medium, counted,and plated onto the chick cardiomyocyte cultures at HUVECto cardiomyocyte ratios of 1:4 or 1:10 and cultured for 6 d.Medium was changed at day 3.

2.2. Cre/lox reporter system

An adenovirus-based Cre/lox reporter system (MicrobixBiosystems, Ontario, Canada) was used to detect fusion be-tween cardiomyocytes and HUVECs. In this system, one ade-novirus mediates expression of Cre recombinase under controlof the human CMV promoter (AdCre). The reporter adenovirusencodes a floxed lacZ cassette, composed of the human CMVpromoter, a spacer sequence flanked by loxP sites, and the lacZcDNA (AdfloxlacZ). The spacer sequence prevents expressionof lacZ, and excision of the spacer is dependent on Cre recom-binase activity. E1A expressing 293 cells [20] were infectedwith AdCre or AdfloxlacZ. The viruses were then propagated,purified and concentrated as previously described [20]. Adeno-virus titer was determined by spectrophotometry at 260 nm(OD 1 ≈ 1012 particles/ml). Adherent monocultures of chickcardiomyocytes and HUVECs were infected with AdCre andAdfloxlacZ, respectively, at 250 virus particles per cell for4 h. After infection, cultures were washed 8 times with PBS.Cells were then trypsinized, pelleted by centrifugation and re-suspended in 10 ml PBS. Centrifugation and resuspensionsteps were repeated twice and the final pellets were resus-pended in 1 ml Medium 200, containing LSGS. Infected cardi-omyocytes and HUVECs were then co-cultured at a ratio of10:1. β-galactosidase was detected in transduced cells usingX-gal staining. Cells were fixed in 2% formaldehyde, 0.2%glutaraldehyde in PBS for 10 minutes at room temperature.They were then rinsed with PBS, and stained for 16 h at37 °C with X-gal (0.1% 5-bromo-4-chloro-3-indolyl D-galacto-side in 5 mM potassium ferricyanide, 5 mM potassium ferro-cyanide, 2 mM MgCl2). Controls included X-gal staining ofmonocultures of each cell type and AdfloxlacZ infected HU-VECs, as a test for Cre-independent lacZ expression as well as

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528522

monocultures of AdCre infected chick cardiomyocytes as a testfor cell viability. In addition the last of the 8 successive rinsesof chick cardiomyocytes following the 4 h adenovirus infectionperiod (washed with medium 200 instead of PBS) was used asa culture media for AdfloxlacZ infected HUVECs to determinewhether any cells became lacZ-positive. This tested whetherresidual AdCre virus that might be released from cardiomyo-cytes following the infection was capable of infecting HU-VECs in the co-culture assay system. No lacZ-positive cellswere observed.

In the adenovirus studies we found that nearly 100% of theHUVECs could be transduced while only 30% of the chickcardiomyocytes could be transduced without toxic effects tothe cells. The fact that 70% of the chick cardiomyocytes werethus not carrying AdCre means that this experimental systemunderestimates the percentage of spontaneous HUVEC-cardio-myocyte heterokaryons that actually occur.

2.3. Indirect immunofluorescence microscopy

Cultured cells were rinsed twice with PBS, fixed with 2%paraformaldehyde (PFA) in PBS for 10 min and permeabilizedwith 1% Triton X-100 in PBS for 10 min. After washing inPBS, cultures were blocked with 1% BSA, 0.05% Tween-20in PBS (blocking solution) for 20 min and incubated with pri-mary antibodies for 1 to 2 h at RT. After thorough washingwith blocking solution, the cultures were incubated with sec-ondary antibodies and washed again with blocking solution for0.5 to 1 h at RT. After a final wash with PBS, the specimenswere rinsed in distilled water and mounted in gelvatol (AirProducts) with 100 mg/ml DABCO (Sigma). Indirect immuno-fluorescence microscopy was performed on a Zeiss Axioplanmicroscope using plan-neofluar 40X (N.A. = 0.73) and 100X(N.A. = 1.30) objectives. Images were acquired using aDAGE-MTI 3CCD camera and imported into Adobe Photo-shop using the Scion Series 7 import plug-in module (V. 1.4).

2.4. Antibodies

Anti-sarcomeric myosin heavy chain monoclonal (mAbs)MF20 (gift of D.A. Fischman, Cornell U. Med. Col.) andF59 (gift of Frank Stockdale, Stanford U.) were used at dilu-tions of 1:100 and 1:10, respectively. Endothelial cells wereidentified using anti-von Willebrand Factor (vWF) polyclonalantibody (pAb) (Sigma, St. Louis, MO) at 1:200 dilutions. Hu-man endothelial cells were distinguished from chick and ratcells in co-culture experiments by immunostaining with anti-human nuclear antigen (HNA) mAb (Chemicon International,Temecula, CA) at a dilution of 1:100. Fused cells were identi-fied using an anti β-galactosidase pAb (Cortex Biochem, SanLeandro, CA) at a dilution of 1:2,000. Secondary antibodiesconjugated to Alexa fluorophores (Molecular Probes) wereused at dilutions of 1:100. For samples reacted with bothMF20 (mouse IgG2b mAb) and anti-HNA mAb (IgG1), anti-mouse IgG2b antibody conjugated to Alexa 594 and anti-

mouse IgG1 Alexa 488 or Alexa 350 were used as secondaryantibodies.

2.5. RT-PCR

Total RNA was isolated from control hearts and culturedcells, and treated with RNase-free DNase I, amplification grade(Invitrogen, Carlsbad, CA). 2 μg total RNA was reverse tran-scribed in a volume of 80 μl as previously described [21]. As anegative control, reverse transcriptase was omitted in parallelreactions. For Real-Time PCR, 2 μl cDNA was added to 18 μlmaster mix, for a final concentration of 1x SYBR Green (Ap-plied Biosystems, Foster City, CA) and 300 μM primers. Allprimers were designed to be human specific. Triplicates of thecDNA were amplified for 40 cycles with the Opticon I Real-Time thermal cycler (MJ Research, Inc., Waltham, MA). ThePCR product level is expressed as Ct, the amplification cycle atwhich the emission intensity of the product rises above an ar-bitrary threshold level. This experiment was performed onetime. A similar experiment in which the cDNAs were subjectedto two successive rounds of 30 PCR cycles, using nested pri-mers, gave similar results. The human primers used in thisstudy are GAPDH forward: ccagcaagagcacaagagga, GAPDHreverse: gagcacagggtactttattgatgg, β-MyHC forward: ccaacaccaacctgtccaag, β-MyHC reverse: cttcctcccaaggagctgtt, cTnIforward: cctgcggagagtgaggatct, cTnI reverse: gaggaataaagcttctctc, cTnI nested forward: ccgggctaaggagtccct, cTnInested reverse: gagctgagccttcctgccta.

2.6. Calculation of relative expression levels of human cardiacgenes in HUVEC-chick cardiomyocyte co-cultures

Human cardiac muscle gene expression in HUVEC-chickcardiomyocyte co-cultures could be attributed to expressionwithin HUVEC-cardiomyocyte heterokaryons, to expressionwithin non-fused “transdifferentiated” HUVECs in responseto cardiomyocyte signals, or to both phenomena. (The calcula-tions below relate to Table 1). To determine the relative levelof human cardiac muscle gene expression, we compared Ctvalues for cardiac troponin I (cTnI) from co-cultured and adulthuman heart tissue. At an initial 4:1 ratio of chick cardiomyo-cytes to HUVECs, the ratios per 1000 total cells at t0 would be800 chick cardiomyocytes and 200 HUVECs. Our growth stu-dies of HUVECs show that they double every 30 h. Thus after6 d (144 h) in culture the HUVECs undergo 4.8 doublings.Prior to co-culture, the cardiomyocytes are treated with mito-mycin C and are therefore non-proliferative. After 6 d and 4.8HUVEC doublings, the initial 1000 cells would theoreticallyyield 5,572 HUVECs and 800 chick cardiomyocytes, a totalof 6,372 cells. If 0.15% of HUVECs are lacZ+ (see Resultssection showing that this fraction of chick embryo cardiomyo-cytes spontaneously fuse with HUVECs in culture) then 9 cellsper 5,572 HUVECs are heterokaryons. However, only about30% of chick heart cells were successfully transduced withCre recombinase (see Methods section: Cre/lox Reporter Sys-tem). Thus if 0.15% HUVECs exhibited lacZ at the end of the

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528 523

experiment, an additional 0.3% of the HUVECs would be un-detectable heterokaryons (due to their lack of Cre recombi-nase); therefore the calculated number of heterokaryons per5,572 HUVECs is 25 or 1 heterokaryon per 233 HUVECs.Only 16 PCR cycles are needed to detect cTnI transcripts inthe adult heart. Thus with a “per cell” ratio of heterokaryonsto total cells of 1:233, the additional PCR cycles that would berequired to detect adult heart levels of cTnI transcripts in het-erokaryons, assuming near perfect PCR efficiency, would beapproximately 8, thus requiring 24 total cycles for detection.To detect cTnI transcripts in heterokaryons at levels 0.1% ofadult levels would require an additional 10 cycles (34 total),which is still 3 cycles below the “noise expression level” inHUVECs alone. Therefore, even if no transdifferentiation oc-curred and all the human cTnI expression in co-cultures weredue solely to HUVEC-chick cardiomyocyte heterokaryon, thelevel of expression by each heterokaryon HUVEC nucleus isless than 0.1% that in the human heart.

2.7. Statistics

The number of cells exhibiting a "dual cardiomyocyte-en-dothelial cell phenotype" is expressed as the percent myosin+

cells that react with Ab against vWF or HNA. Mono- or multi-nucleated dual phenotype cells are reported as a percentage ofcells with both cardiomyocyte and endothelial cell markerscontaining one, or more than one, DAPI stained nucleus. Datais reported as means ± one standard deviation.

3. Results

3.1. Most HNA+/myosin+ cells in HUVEC-rat neonatalcardiomyocyte co-cultures are multinucleated

It was previously reported that HUVECs express cardiacmarkers when co-cultured with rat neonatal cardiomyocytes[3]. This was interpreted as being due to transdifferentiationof HUVECs in response to cardiomyocyte signals. In an at-tempt to extend these studies we isolated rat neonatal cardio-myocytes and cultured them with HUVECs at a ratio of 10:1for 6 d. The cultures were then fixed and immunostained withMF20, a mAb that reacts with both rat and human sarcomericmyosin and an antibody specific for human nuclear antigen(HNA) that distinguishes between HUVECs and rat cardio-myocytes in co-cultures. Examination of the immunostainedco-cultures indicated that HNA+/myosin+ cells represented1.2 ± 0.5% of the total HUVEC population (3,570 HNA+ cellscounted on hundreds of random fields in 5 culture plates). Im-portantly, within the very small population of HNA+/myosin+

cells, 67 ± 15.5% were multinucleated. When cultured alone,only 13% of the HUVECs are multinucleated, suggesting thatmany of the HNA+/myosin+ cells could be the result of fusion;or that multinucleated HUVECs are disproportionately suscep-tible to cardiomyocyte conversion.

3.2. Most HNA+/myosin+ cells in HUVEC-E3 chickcardiomyocyte co-cultures are also multinucleated

Since the mononucleated HUVECs in rat cardiomyocyte co-cultures exhibited little indication of conversion to a cardiacphenotype, we examined the possibility that co-culturingHUVECs with developmentally earlier cardiac muscle cellsmight provide an environment more conducive for phenotypicconversion. This possibility seemed plausible because cardio-myocytes and cardiac endothelial cells may be derived fromsimilar precursor cells [10,11,22]. If cell-cell interactions indeveloping hearts influence emergence of the cardiomyocytephenotype, it seemed possible that co-culturing HUVECs withearly embryonic heart cells might induce their phenotypic con-version.

The ability of E3 chick heart cells to induce HUVECs toexhibit a cardiac phenotype was tested at a 10:1 ratio of heartcells to HUVECs. At this stage chick hearts contain 90% cellsthat are immuno-positive for sarcomeric myosin heavy chain,6% cells that are immuno-positive for the endothelial cell mar-ker vWF and 4% cells that are presumed to be fibroblasts [9].After six days of co-culture, HUVECs, cells with green (HNA-positive) nuclei (Fig. 1 A and B, arrows), could be distin-guished from chick cardiomyocytes, cells with red cytoplasmand unstained nuclei (Fig. 1 A-D, arrowhead). Cells exhibitingboth HNA and myosin expression (Fig. 1B and D, arrows)were presumed to be of HUVEC origin. However, the totalpercentage of HNA+ cells exhibiting myosin was only0.4% ± 0.2 based on counting HNA+/MyHC+ cells in hundredsof fields at 400X magnification. While these data would beconsistent with the possibility that HUVECs can express cardi-ac genes when co-cultured with the total cell population fromembryonic chick hearts, data collected from three separate ex-periments show that about 75% of these cells are multi-nu-cleated (Fig. 1D), which again raises the possibility that someor all of HNA+/myosin+ cells result from HUVEC-cardiomyo-cyte fusion (this is tested directly in studies described below.However, almost all the HNA+/myosin+ cells (Fig. 1B and D)contained more than one HNA+ nucleus. While it would beanticipated that a heterokaryons formed by the fusion of oneHUVEC and one chick cardiomyocyte would contain oneHNA+ nucleus and one HNA- nucleus originating from the hu-man and chick cell respectively, nearly all of the binucleatedHNA+/myosin+ cells contained two HNA+ nuclei (Fig. 1D).This reflects the ability of human nuclear antigens to localizeto chick as well as human nuclei in chick-human heterokar-yons. This observation is consistent with a previous study thatanalyzed the reactivation of chick erythrocyte nuclei in chickerythrocyte-human tumor (HELA) heterokaryons [23]. Thisstudy showed that 42 hours after fusion, close to 100% ofchick nuclei in chick-human heterokaryons reacted positivelywith anti human nucleolar specific antibody, and over 80%were reactive with anti human nucleoplasm antibody. In addi-tion, HeLa cell nuclei in the heterokaryons reacted positivelywith an antibody specific for chick nucleolar antigens. This andan analogous study with chick erythrocyte rat myoblast hetero-

Fig. 1. HUVECs co-cultured with chick embryonic myocytes exhibit cardiac markers. HUVECs were co-cultured with chick heart cells for 6 d at a ratio of 1:10. Thecells were then fixed, stained with DAPI (blue) and probed with Abs against HNA (green), and myosin (red). Myosin expressing cells that contain DAPI+/HNA-

nuclei (arrowheads) are chick cardiomyocytes. Cells exhibiting myosin and HNA+ nuclei (arrows) are HUVECs, but they may also be heterokaryons (see also Fig. 2)in which HNA has entered the chick cardiomyocyte nucleus.

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528524

karyons [24] directly demonstrate that mammalian nuclear anti-gens are transported into chick nuclei within heterokaryons.

An explanation for why 25% of the HNA+/myosin+ cells(0.1% of the entire HUVEC population) are mono-nucleatedis more problematic. Since no HNA+/myosin+ cells were de-tected among comparable numbers of HUVECs cultured bythemselves, the approximately one mononucleatedHNA+/myosin+ cell per 1000 HUVECs in co-cultures implieseither that only very rare HUVECs have the developmentalcapacity to respond to an embryonic heart cell-mediated induc-tive environment; or, that the inductive process has a criticalconcentration threshold or cell-cell contact requirement, suchthat only about one HUVEC per thousand experiences the in-ductive event. Alternatively, the rare mononucleatedHNA+/myosin+ cells could be due to HUVEC-cardiomyocytecell fusions in which one nucleus was lost or in which theoriginal heterokaryon nuclei fused.

To test whether E3 cardiomyocytes could induce otherchick endothelial cells to convert to a cardiomyocyte pheno-type, E3 heart cells were co-cultured with embryonic (E10-14) chick aorta endothelial cells; no sarcomeric myosin wasdetected among hundreds of thousands of chick aortic endothe-lial cells (data not shown).

3.3. Chick embryo cardiomyocytes spontaneously fuse withHUVECs in culture

To provide an unambiguous test of whether HUVECs fusewith E3 cardiomyocytes, HUVECs were transduced with ade-noviruses that contained a floxed lacZ cassette and E3 chickheart cells were transduced with a Cre-recombinase expressioncassette (Fig. 2 A). Fusion of these infected cells should result

in Cre-directed recombination of the lacZ cassette causing ex-cision of the spacer region and expression of lacZ, detectableby X-Gal staining or immunofluorescence histochemistry.However, since only about 30% of the chick cardiomyocytescould be stably transduced (see Materials and Methods), thepercentage of heterokaryons detected underestimates the truefrequency of spontaneous HUVEC-cardiomyocyte fusion by70% (see below).

Four hours following adenovirus addition, HUVECs andcardiomyocytes were dissociated, washed extensively andmixed at a ratio of 1:10. After six days the cells were fixedand stained with X-gal (Fig. 2B) or probed with antibodiesagainst HNA (Fig. 2C), sarcomeric myosin (Fig. 2D) and β-galactosidase (lacZ) (Fig. 2E). X-gal staining allowed the iden-tification of cardiomyocyte-endothelial fused cells in the Cre/lox experiments and among these lacZ+ cells we also foundwell-organized myofibrils (Fig. 2B). Cells containing all threemarkers were identified (Fig. 2C-F) indicating that the presenceof sarcomeric myosin in cells containing the HNA marker isdue to the fusion of HUVECs and E3 chick cardiomyocytes;and significantly, most of the triply labeled cells were multi-nucleated (see Fig. 2B, and below). Among the total HUVECsscored, the 763 lacZ+ cells counted on 4 entire culture dishsurfaces, we calculated that these represented slighty less than0.15% of an estimated 5 x 105 total cells (based on counting allcells in 10 random fields in each of the 4 dishes).

In these studies it was necessary to control for the detectionof false-positive cell fusion due to the possibility of residualAdCre virus from chick cardiomyocyte-transduced culturessecondarily infecting AdfloxlacZ-transduced HUVECs. Whilethe residual AdCre virus hypothesis seemed unlikely due to thefact that E3 heart cells were rinsed 8 times following the 4 h

Fig. 2. Myosin-positive cells containing human nuclei result from cell fusion. A) A summary of Cre/lox recombination technique. Cardiomyocytes were infectedwith AdCre and HUVEC cultures were infected with AdfloxlacZ. B-F) Fusion between HUVECs and cardiomyocytes was demonstrated within cells in which thespacer region of AdfloxedlacZ was excised, resulting in an active form of β-gal. 4 h after infections the cardiomyocytes and HUVECs were rinsed 8 times and thentrypsinized, washed once, and co-cultured at a ratio of 10:1 for 6 d (Panels B-F). The cells were fixed and stained with X-gal (B) or with anti-HNA followed by a goatanti-mouse Alexa-350 conjugated secondary antibody (C) and reacted with abs specific for sarcomeric myosin (D) and β- gal E). Fused cells were identified as beingblue (B) when stained with X-gal and observed under phase contrast microscopy, or as β-gal+ (E and F) when observed using indirect immunofluorescence. The cellshown in C-F is the product of such fusion and it is positive for HNA and myosin.

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528 525

virus infection, and then-dissociated, centrifuged, resuspendedto 5 ml, recentrifuged and resuspended prior to addition toHUVEC cultures, this was tested by culturing AdfloxlacZ-in-fected HUVECs in the final rinse media used to remove resi-dual AdCre virus from chick cardiomyocyte cultures. No β-galactosidase+ cells were observed in these cultures. This de-monstrated that by the end of the rinsing process no residualAdCre virus remained in the transduced E3 heart cultures thatwould have been capable of infecting HUVECs in the co-cul-ture assay system. However, since the chick cardiomyocyteswere then trypsin-dissociated, centrifuged, resuspended and re-centrifuged, the negative data does not totally eliminate thepossibility that some β-galactosidase+ cells could result from

the release of AdCre virus from dying cells. If infectious Ad-Cre virus were released, it would then, however, be necessaryto account for the fact that the majority of lacZ+ cells also con-tain multiple nuclei; this observation is much more easily ex-plained by HUVEC-cardiomyocyte fusion. Thus in our studieswe believe that essentially all of the lacZ+ cells resulted fromHUVEC-chick cardiomyocyte fusion. Since only about 30% ofthe chick cardiomyocytes were transduced with Cre recombi-nase, thereby making it possible to detect only about one thirdof the actual HUVEC-cardiomyocyte- heterokaryons, the totalcalculated percentage of heterokaryons was about 0.5%.

Data acquired from three separate experiments also revealedthat of the total lacZ+ cells, 13.8 ± 6.1% were mono-nucleated

Table 1RT-Real-Time-PCR with human primers

TISSUE/ CELLS PCR Targets and Cycle Threshold (Ct)a

GAPDH β-MyHC Cardiac TnIChick Heart (E3) no product no product no product

@40 cyclesChick Cardiomyocytes (cCMCs) in vitro (6 d) no product no product not analyzedHuman Heart (biopsy) 17 cycles 15b 16b

HUVECs in vitro (6 d) 15 no product 37cCMCs/ HUVECs (4:1) Co-Culture (6 d) 17 38 36cCMCs/ HUVECs (10:1) Co-Culture (6 d) 17 38b 37b

a Cycle Threshold (Ct): the amplification cycle at which the emission intensity of the product rises above an arbitrary threshold level.b β-MyHC and cTnI Cts from human heart were defined as 1 for comparisons to β-MyHC and cTnI mRNA levels calculated for each cell culture type. Even

at 10:1 ratios of chicken cardiomyocytes to HUVECs, β-MyHC and cTnI transcripts were at least one hundred thousandth lower than in adult heart.

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528526

while 86.2 ± 6.1% were multi-nucleated. Since the fraction ofmono-nucleated fused cells in these experiments is in the samerange as the roughly 25% mono-nucleated HNA+/myosin+

cells observed in the experiments depicted in Fig. 1, it seemslikely that most of the mono-nucleated cells containing bothHNA and sarcomeric myosin result from original heterokar-yons that have lost one or more nuclei, and/or in which thetwo nuclei have also fused. Thus among the very low numberof “HUVECs” exhibiting myosin+ immunostaining, the vastmajority can be attributed to HUVEC-cardiomyocyte fusionrather than to cardiomyocyte-mediated induction of HUVECsto activate human cardiac gene expression.

3.4. Human cardiomyocyte genes are not activatedin HUVEC-cardiomyocyte co-cultures

The preceding studies indicate that most, if not all, dualphenotype cells result from HUVEC-cardiomyocyte fusion;but they do not address whether fusion leads to the activationof cardiac muscle gene expression within HUVEC-derived nu-clei. Furthermore, if some HUVECs do indeed transdifferenti-ate to a cardiomyocyte phenotype, the extent to which humancardiac muscle genes are expressed was not determined. Thesepossibilities were examined in 6-day HUVEC-E3 chick heartco-cultures via quantitative RT-PCR with primers specific forhuman genes (cDNA from a human heart biopsy was used as apositive control for all the human primer sets). We tested fortwo cardiomyocyte genes, β-MyHC and cardiac TnI (cTnI);and human GAPDH served as a control for the recovery andamplification of human mRNAs. β-MyHC mRNA is the pre-dominant MyHC isoform expressed in fetal and adult ventri-cles. In the adult, it is also found in the atria and slow musclefibers [25]. cTnI is expressed exclusively in the heart and inembryonic, fetal and adult human stages [26,27].

In the 6-day co-culture sample, human GAPDH, β-MyHCand cTnI mRNA are detectable after 17 cycles (Table 1). As anegative control, cDNA from embryonic chick heart and cul-tured chick cardiomyocytes was employed. No product wasdetected after 40 PCR cycles with the chick cDNA alone usinghuman-specific primers. When HUVEC cDNA was tested

using primers specific for human sarcomeric proteins no pro-duct was detectable until 37 cycles, whereas human GAPDHtranscripts were detected after 15 cycles. Products detectedafter 36 PCR cycles are considered to be in the “noise range”of the technique and probably do not represent amplificationfrom biologically relevant numbers of transcripts per cell.When cDNA from co-cultures containing either 4:1 or 10:1ratios of cardiomyocytes to HUVECs was tested, β-MyHC orcTnI amplimers were not detected until 36 cycles. Taking intoaccount the > 20 PCR cycle difference in Ct values betweenthe heart biopsy and co-culture samples, we estimate that theamounts of human β-MyHC and cardiac TnI transcripts perhuman GAPDH transcript are at least one hundred thousandththe amounts present in adult human cardiomyocytes. It thusseems unlikely that appreciable numbers of HUVECs are in-duced to transcribe significant levels of cardiac muscle genesin response to co-culture with embryonic cardiomyocytes.Furthermore, it appears that human cardiac muscle gene ex-pression was not activated in the HUVEC heterokaryons thatrepresent about 0.5% of the total cells in such co-cultures, evento levels equivalent to about 0.1% that of cTnI gene expressionin adult human heart tissue.

4. Discussion

The possible conversion of HUVECs to cardiomyocytes byco-culture with newborn rat cardiomyocytes was suggested bya previous study [3]. Analogous studies with endothelial pre-cursor cells (EPCs) derived from human bone marrow sug-gested that as many as 10% of these cells convert to cardio-myocytes when co-cultured with rat neonatal cardiomyocytes[5]. In addition, conversion of circulating endothelial precursorcells to cardiomyocytes in co-cultures with neonatal rat cardi-omyocytes was reported to be potentiated by exposure to non-canonical Wnt 11 [4]. However, because the extent of cell fu-sion was not determined in any of these studies, fusion mightwell account for phenotypes interpreted as being due to trans-differentiation.

The initial goal of our study was to improve the methods forconversion of HUVECs to the cardiomyocyte lineage, and

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528 527

hopefully to discover the cardiomyocyte-mediated factors in-volved. In our experimental system we tested the ability ofboth neonatal rat and early embryonic chick heart cells to causeHUVECs to transdifferentiate. Our results with neonatal rat co-cultures differ from those of Condorelli et al. [3], in that wefound no clear evidence of the transdifferentiation of HUVECsinto cardiomyocytes. Rather we found that virtually all HU-VEC-derived cells with dual endothelial and cardiomyocytephenotypes could be attributed to HUVEC-cardiomyocyte cellfusion. We also obtained similar data in HUVEC co-cultures inwhich the cardiomyocytes were derived from early chick em-bryos.

Condorelli et al. (2001) also reported that about the samepercentage of mouse embryo aorta endothelial cells becometransdifferentiated when co-cultured with newborn rat cardio-myocytes as did HUVECs. We examined this possibility in co-cultures of E3 chick heart cells and E10-14 chick aorta en-dothelial cells that had been pre-labeled with acetylated lowdensity lipoprotein conjugated to Alexa 488, or with greenfluorescent cell tracker (Molecular Probes, Eugene, OR), andfound no evidence for transdifferentiation of the endothelialcells. This negative data partially addresses the possibility thatthe failure of E3 chick cardiac cells to induce HUVEC differ-entiation to a cardiomyocyte phenotype is due to chick-humaninterspecies incompatibility that does not exist between rat andhuman cells. However, even with a complete species match, noendothelial cell transdifferentiation occurred.

In our studies we used a Cre/lox reporter strategy to accessfusion between HUVECs and cardiomyocytes. The HUVECshad been transduced with adenovirus carrying a floxed lacZcassette composed of a CMV promoter, a spacer sequenceflanked by lox P sites, and the lacZ cDNA; the E3 chick heartshad been transduced with adenovirus carrying a CMV-drivenCre recombinase cDNA. LacZ-positive cells presumably result-ing from HUVEC-chick cardiomyocyte fusion were observed,and significantly, most of these contained 2 nuclei (Fig. 2B,F).Thus even if a few of the lacZ-positive cells were due to resi-dual AdCre virus released from chick heart cells, or to a pos-sible viral transfer to HUVECs via intercellular connections[28], this would not explain why 86% of the lacZ-positive cellswere multinucleated. As an additional control, AdfloxlacZ-transduced HUVECs were cultured alone for 6 days and thenstained for lacZ as a test for either leaky activation of theCMVfloxlacZ cassette or for endogenous β-galactosidase ex-pression by senescing HUVECs [29,30]. No lacZ-positive cellswere observed. Taken together we believe that our results canonly be due to HUVEC-cardiomyocyte fusion. Rare in vivoinstances of cardiomyocyte fusion have also been reported inthe hearts of mice injected with bone marrow derived cells[15], and skeletal muscle cells [17].

Although we observe myosin in all lacZ+ cells, we do notfind appreciable expression of human cardiac specific genes inthe HUVEC-chick cardiomyocyte co-cultures (Table 1). Thisindicates that HUVECs subjected to putative cardiogenic in-ductive effects from surrounding cardiomyocytes do not trans-differentiate into cardiomyocytes. In addition, HUVEC nucleiwithin HUVEC-cardiomyocyte heterokaryons do not initiate a

cardiomyocyte gene program despite the presence of chick car-diomyocyte transcription factors.

While our study was nearing completion, a partially analo-gous report appeared in which HUVECs, cardiac fibroblasts orbone marrow-derived cells were co-cultured with neonatal ratcardiomyocytes and found to spontaneously fuse with the car-diomyocytes [18]. Consistent with our observations, the inves-tigators found the percentage of dual phenotype cells in co-cultures of endothelial cells and cardiomyocytes to be far lower(~0.05% vs. ~10%) than reported by previous investigators [3,5]. In addition, they observed that the number of HUVEC-car-diomyocyte heterokaryons expressing an endothelial specificmarker (vWF) decreased over time. This was interpreted asimplying that in such heterokaryons, the cardiomyocyte pheno-type is dominant. However, these heterokaryons were nottested for human cardiac gene expression, thus it remained un-known whether fusion between HUVECs and cardiomyocytescould induce the HUVEC nuclei to activate transcription ofcardiomyocyte-specific genes. We used quantitative RT-PCRto examine whether HUVEC nuclei in co-cultures containingabout 0.5% heterokaryons are “reprogrammed” to activate acardiomyocyte gene program. In the case of the HUVEC-chickcardiomyocyte-co-cultures we observed no evidence of humancardiac gene transcription, despite the fact that the RT-PCRanalysis would have detected human MyHC transcripts at le-vels as low 0.01% that in adult cardiomyocytes, even if only0.5% of the HUVECs had transdifferentiated or been in hetero-karyons. This very low limit to the possible level of humancardiac gene activation suggests that neither the HUVEC het-erokaryon nuclei, nor the nuclei in HUVECs surrounded byembryonic heart cells, activated detectable levels of cardiacmuscle gene expression. These results are consistent with thoseof Evans et al. (1994) who observed inhibition of cardiac pro-moter function within cardiomyocyte-fibroblast heterokaryons[31].

Based on our studies and those of Matsuura et al. [18] wedo not consider HUVECs as a promising source of cardiomyo-cytes for repair of damaged hearts by injection and proximityinduced transdifferentiation. This conclusion does not precludethe possibility of converting HUVECs to cardiomyocytes viathe forced expression of cardiac muscle transcription factorsor other inducers. However, if strategies of this type were dis-covered, they might be more effectively applied toward pa-tient-derived circulating endothelial cell precursors, as thesecells may be more amendable to conversion [4].

Acknowledgments

We thank J. Angelo and J. Buskin for insightful discussionsand for critical comments on the manuscript. This work wassupported by National Institutes of Health grant HL064387(BioEngineered Autologous Tissue (BEAT) Grant) and post-doctoral training grants from the American Heart Association(award # 0120630Z) and Cardiovascular Research TrainingProgram at the University of Washington.

[18

[19

[20

[21

[22

[23

[24

[25

[26

[27

[28

[29

[30

[31

R.E. Welikson et al. / Journal of Molecular and Cellular Cardiology 40 (2006) 520–528528

References

[1] Orlic D, et al. Bone marrow cells regenerate infarcted myocardium. Nat-ure 2001;410(6829):701–5.

[2] Orlic D, et al. Mobilized bone marrow cells repair the infarcted heart,improving function and survival. Proc Natl Acad Sci USA 2001;98(18):10344–9.

[3] Condorelli G, et al. Cardiomyocytes induce endothelial cells to trans-dif-ferentiate into cardiac muscle: implications for myocardium regeneration.Proc Natl Acad Sci USA 2001;98(19):10733–8.

[4] Koyanagi M, et al. Non-canonical Wnt Signaling Enhances Differentia-tion of Human Circulating Progenitor Cells to Cardiomyogenic Cells. JBiol Chem 2005;280(17):16838–42.

[5] Badorff C, et al. Transdifferentiation of blood-derived human adult en-dothelial progenitor cells into functionally active cardiomyocytes. Circu-lation 2003;107(7):1024–32.

[6] Urbanek K, et al. Intense myocyte formation from cardiac stem cells inhuman cardiac hypertrophy. Proc Natl Acad Sci USA 2003;100(18):10440–5.

[7] Oh H, et al. Cardiac progenitor cells from adult myocardium: homing,differentiation, and fusion after infarction. Proc Natl Acad Sci USA2003;100(21):12313–8.

[8] Laugwitz KL, et al. Postnatal isl1+ cardioblasts enter fully differentiatedcardiomyocyte lineages. Nature 2005;433(7026):647–53.

[9] Welikson RE, Kaestner S, Hauschka SD. Some Embryonic Cardiomyo-cytes Express Endothelial Genes. In preparation 2005.

[10] Linask KK, Lash JW. Early heart development: dynamics of endocardialcell sorting suggests a common origin with cardiomyocytes. Dev Dyn1993;196(1):62–9.

[11] Eisenberg CA, Bader D. QCE-6: a clonal cell line with cardiac myogenicand endothelial cell potentials. Dev Biol 1995;167(2):469–81.

[12] Terada N, et al. Bone marrow cells adopt the phenotype of other cells byspontaneous cell fusion. Nature 2002;416(6880):542–5.

[13] Vassilopoulos G, Wang PR, Russell DW. Transplanted bone marrow re-generates liver by cell fusion. Nature 2003;422(6934):901–4.

[14] Wang X, et al. Cell fusion is the principal source of bone-marrow-de-rived hepatocytes. Nature 2003;422(6934):897–901.

[15] Alvarez-Dolado M, et al. Fusion of bone-marrow-derived cells with Pur-kinje neurons, cardiomyocytes and hepatocytes. Nature 2003;425(6961):968–73.

[16] Beltrami AP, et al. Adult cardiac stem cells are multipotent and supportmyocardial regeneration. Cell 2003;114(6):763–76.

[17] Reinecke H, Minami E, Poppa V, Murry CE. Evidence for fusion be-tween cardiac and skeletal muscle cells. Circ Res 2004;94(6):e56–e60.

] Matsuura K, et al. Cardiomyocytes fuse with surrounding noncardiomyo-cytes and reenter the cell cycle. J Cell Biol 2004;167(2):351–63.

] Iwaki K, Sukhatme VP, Shubeita HE, Chien KR. Alpha- and beta-adre-nergic stimulation induces distinct patterns of immediate early gene ex-pression in neonatal rat myocardial cells. fos/jun expression is associatedwith sarcomere assembly; Egr-1 induction is primarily an alpha 1-mediated response. J Biol Chem 1990;265(23):13809–17.

] Murry CE, Kay MA, Bartosek T, Hauschka SD, Schwartz SM. Muscledifferentiation during repair of myocardial necrosis in rats via gene trans-fer with MyoD. J Clin Invest 1996;98(10):2209–17.

] Kastner S, Elias MC, Rivera AJ, Yablonka-Reuveni Z. Gene expressionpatterns of the fibroblast growth factors and their receptors during myo-genesis of rat satellite cells. J Histochem Cytochem 2000;48(8):1079–96.

] Sugi Y, Markwald RR. Formation and early morphogenesis of endocar-dial endothelial precursor cells and the role of endoderm. Dev Biol J1–DB 1996;175(1):66–83.

] Ringertz NR, Carlsson SA, Ege T, Bolund L. Detection of human andchick nuclear antigens in nuclei of chick erythrocytes during reactivationin heterokaryons with HeLa cells. Proc Natl Acad Sci USA 1971;68(12):3228–32.

] Bergman M, Nyman U, Ringertz N, Pettersson I. Appearance and originof snRNP antigens in chick erythrocyte nuclei reactivated in heterokar-yons. J Cell Sci 1990;95(Pt 3):361–70.

] Kurabayashi M, Komuro I, Tsuchimochi H, Takaku F, Yazaki Y. Mole-cular cloning and characterization of human atrial and ventricular myosinalkali light chain cDNA clones. JBC 1988;263:13930–6.

] Bhavsar PK, Dhoot GK, Cumming DV, Butler-Browne GS,Yacoub MH, Barton PJ. Developmental expression of troponin I iso-forms in fetal human heart. FEBS Lett 1991;292(1–2):5–8.

] Sasse S, et al. Troponin I gene expression during human cardiac devel-opment and in end-stage heart failure. Circ Res 1993;72(5):932–8.

] Koyanagi M, Brandes RP, Haendeler J, Zeiher AM, Dimmeler S. Cell-to-cell connection of endothelial progenitor cells with cardiac myocytes bynanotubes: a novel mechanism for cell fate changes? Circ Res 2005;96(10):1039–41.

] Kalashnik L, et al. A cell kinetic analysis of human umbilical vein en-dothelial cells. Mech Ageing Dev 2000;120(1-3):23–32.

] Dimri GP, et al. A biomarker that identifies senescent human cells inculture and in aging skin in vivo. Proc Natl Acad Sci USA 1995;92(20):9363–7.

] Evans SM, Tai LJ, Tan VP, Newton CB, Chien KR. Heterokaryons ofcardiac myocytes and fibroblasts reveal the lack of dominance of the car-diac muscle phenotype. Mol Cell Biol 1994;14(6):4269–79.