Retraction

MEDICAL SCIENCESRetraction for “Identification of a coronary vascular progenitorcell in the human heart,” by Claudia Bearzi, Annarosa Leri,Francesco Lo Monaco, Marcello Rota, Arantxa Gonzalez, ToruHosoda, Martino Pepe, Khaled Qanud, Caroline Ojaimi, SilvanaBardelli, Domenico D’Amario, David A. D’Alessandro, Robert E.Michler, Stefanie Dimmeler, Andreas M. Zeiher, Konrad Urbanek,Thomas H. Hintze, Jan Kajstura, and Piero Anversa, which was firstpublished August 27, 2009; 10.1073/pnas.0907622106 (Proc. Natl.Acad. Sci. U.S.A. 106, 15885–15890).The editors wish to note that, based on the recommendation of

Harvard Medical School (HMS) and Brigham and Woman’sHospital (BWH), we are retracting this paper. The HMS andBWH review of images contained in publications from the P.A.laboratory suggests that image data appeared to be modified inways determined to fall outside acceptable image adjustmentpractices. The HMS/BWH found instances of the following: Fig.1A (Top Left) appears to have areas of c-kit-positive stainingselectively inserted into the image plane; Fig. 1A (Bottom Left)appears to have areas of c-kit-positive staining selectively added,as well as the translocation of cells within the image.

May R. BerenbaumEditor-in-Chief

Published under the PNAS license.

First published September 30, 2019.

www.pnas.org/cgi/doi/10.1073/pnas.1916177116

www.pnas.org PNAS | October 8, 2019 | vol. 116 | no. 41 | 20795

RETR

ACT

ION

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Dow

nloa

ded

by g

uest

on

Aug

ust 2

0, 2

020

Identification of a coronary vascular progenitorcell in the human heartClaudia Bearzia, Annarosa Leria, Francesco Lo Monacob, Marcello Rotaa, Arantxa Gonzaleza, Toru Hosodaa,Martino Pepeb, Khaled Qanudb, Caroline Ojaimib, Silvana Bardellia, Domenico D’Amarioa, David A. D’Alessandroc,Robert E. Michlerc, Stefanie Dimmelerd, Andreas M. Zeiherd, Konrad Urbaneka, Thomas H. Hintzeb, Jan Kajsturaa,and Piero Anversaa,1

aDepartments of Anesthesia and Medicine and Cardiovascular Division, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115;bDepartment of Physiology, New York Medical College, Valhalla, NY 10595; cMontefiore Medical Center, Albert Einstein College of Medicine,New York, NY 10467; and dDepartment of Internal Medicine III, J.W. Goethe University, Frankfurt, Germany

Communicated by Eugene Braunwald, Brigham and Women’s Hospital, Boston, MA, July 16, 2009 (received for review April 5, 2009)

Primitive cells capable of generating small resistance arterioles andcapillary structures in the injured myocardium have been identifiedrepeatedly. However, these cells do not form large conductive coro-nary arteries that would have important implications in the manage-ment of the ischemic heart. In the current study, we determinedwhether the human heart possesses a class of progenitor cells thatregulates the growth of endothelial cells (ECs) and smooth musclecells (SMCs) and vasculogenesis. The expression of vascular endothe-lial growth-factor receptor 2 (KDR) was used, together with the stemcell antigen c-kit, to isolate and expand a resident coronary vascularprogenitor cell (VPC) from human myocardial samples. Structurally,vascular niches composed of c-kit-KDR-positive VPCs were identifiedwithin the walls of coronary vessels. The VPCs were connected by gapjunctions to ECs, SMCs, and fibroblasts that operate as supportingcells. In vitro, VPCs were self-renewing and clonogenic and differen-tiated predominantly into ECs and SMCs and partly into cardiomyo-cytes. To establish the functional import of VPCs, a critical stenosiswas created in immunosuppressed dogs, and tagged human VPCswere injected in proximity to the constricted artery. One month later,there was an increase in coronary blood flow (CBF) distal to thestenotic artery, resulting in functional improvement of the ischemicmyocardium. Regenerated large, intermediate, and small humancoronary arteries and capillaries were found. In conclusion, the hu-man heart contains a pool of VPCs that can be implemented clinicallyto form functionally competent coronary vessels and improve CBF inpatients with ischemic cardiomyopathy.

The etiology of ischemic myocardial injury is represented bylesions of the major epicardial coronary arteries that restrict

blood flow to the distal myocardium, leading to infarction and scarformation. Successful intervention would require restoration of theintegrity or replacement of the damaged coronary arteries tointerfere with the etiology of the disease and the development ofthe myopathy (1). This possibility would change dramatically thegoal of cell therapy for the ischemic heart; prevention of myocardialinjury would become the end point of cell therapy rather than thepartial repair of established damage.

Vascular progenitor cells (VPCs) are located in the proepicar-dium from where they migrate into the myocardium and differen-tiate into endothelial cells (ECs) and smooth muscle cells (SMCs)organized in coronary vessels. Myocyte progenitor cells (MPCs)distributed in the cardiac crescent and pharyngeal mesodermdifferentiate and progressively constitute together with the coro-nary vessels the four-chambered heart (2). Here, we report that thehuman heart contains two progenitor cell (PC) classes with inher-ent characteristics that determine their distinct cardiogenic fates.The recognition that VPCs form conductive coronary arteries andthat MPCs form cardiomyocytes supports the notion that these cellsmay be implemented clinically according to the need of the organand uniqueness of the cardiac lesion.

ResultsVascular and Myocardial Niches. If the human heart harbors pheno-typically distinct PC classes, then VPCs and MPCs would be

expected to be nested in different anatomical locations and regu-lated by separate supporting cells (3). We have found vascularniches throughout the coronary circulation and myocardial nichesin the muscle compartment of the organ. Vascular niches, com-posed of clusters of cells expressing c-kit, were identified in epi-cardial coronary arteries, arterioles, and capillaries (Fig. 1A and Fig.S1 A–C in SI Appendix). These cells were positive for KDR, whichis the earliest marker of angioblast precursors and is present inprimitive cells with significant potential for vascular growth (4). Thelack of CD45 and tryptase excluded the contribution of mast cellsto this cell pool. The c-kit-KDR-positive cells were located in theintima, media, and adventitia, and connexin 43 and N-cadherinwere detected at the interface with ECs, SMCs, and fibroblasts,suggesting that these cells may function as supporting cells. We haveconsidered the presence of KDR together with c-kit to be a goodpredictor of VPCs. Similarly, the myocardial interstitium containedniches in which c-kit-positive KDR-negative cells were connectedby junctional proteins to the supporting cells (5), myocytes andfibroblasts (Fig. S1D in SI Appendix). We have assumed theexpression of c-kit in the absence of KDR to be indicative of MPCs.

Human VPCs and MPCs. VPCs and MPCs at P2–P3 were character-ized by FACS. VPCs were negative for hematopoietic markers and�-sarcomeric-actin (�-SA), and showed low levels of the ECadhesion protein CD31 and the SMC marker TGF-�1 receptor(Fig. 1B). MPCs were negative for markers of hematopoietic, EC,and SMC lineages; �-SA was present in a small fraction of cells (Fig.1B). For clonal analysis, VPCs and MPCs were plated in singlewells, and multicellular clones were obtained; clones derived fromVPCs consisted of cells positive for c-kit and KDR, and clonesformed from MPCs contained cells positive for c-kit and negativefor KDR (Fig. 1C). Clonogenic VPCs exposed to differentiatingmedium lost largely c-kit and KDR and expressed transcriptionfactors and cytoplasmic and membrane proteins specific to ECs,Ets1, CD31, and vWf, and SMCs, GATA6 TGF-�1 receptor and�-SMA. Small fractions of cells expressed the myocyte transcriptionfactor MEF2C and the sarcomeric protein �-SA. Differentiatingclonogenic MPCs became predominantly c-kit-negative and ex-pressed Nkx2.5, MEF2C, connexin 43, �-cardiac-actinin, and �-SA.Some cells were positive for Ets1, GATA6, vWf, and �-SMA (Fig.S2A in SI Appendix). The lineage commitment of these cells wasconfirmed by immunocytochemistry (Fig. S2B in SI Appendix).

Author contributions: A.L., M.R., T.H., T.H.H., J.K., and P.A. designed research; C.B., F.L.M.,M.R., A.G., T.H., M.P., K.Q., C.O., S.B., D.D., D.A.D., R.E.M., S.D., A.M.Z., K.U., and J.K.performed research; C.B., A.L., F.L.M., M.R., A.G., T.H., M.P., K.Q., C.O., S.B., D.D., R.E.M.,S.D., A.M.Z., K.U., T.H.H., J.K., and P.A. analyzed data; and A.L., T.H.H., and P.A. wrote thepaper.

The authors declare no conflict of interest.

1To whom correspondence should be addressed. E-mail: panversa@partners.org .

This article contains supporting information online at www.pnas.org/cgi/content/full/0907622106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0907622106 PNAS � September 15, 2009 � vol. 106 � no. 37 � 15885–15890

MED

ICA

LSC

IEN

CES

See Retraction Published September 30, 2019

Collectively, VPCs formed 4.9-fold more SMCs and 5.7-fold moreECs than MPCs, and MPCs formed 6.3-fold more myocytes thanVPCs (Fig. 2A).

Division of VPCs and MPCs and Supporting Cells. Stem cells dividesymmetrically and asymmetrically, and these growth patterns arecontrolled, respectively, by uniform and nonuniform distribution ofthe cell fate determinants Numb and �-adaptin (3). VPCs andMPCs were cultured, and the partitioning of �-adaptin was estab-lished. GATA6, Ets1, and GATA4 were used as markers of cellcommitment. Both VPCs and MPCs divided symmetrically andasymmetrically (Fig. 2 B–H). Asymmetric division of VPCs gave riseto one daughter cell that expressed Ets1 or GATA6, whereas theother failed to show either transcription factor. A comparablebehavior was observed with MPCs.

A functional assay was performed to define the role of connexinsin the formation of gap junctions between VPCs or MPCs, on theone hand, and ECs, SMCs, fibroblasts, and cardiomyocytes, on theother (Figs. S3–S5 in SI Appendix). We established that the humancoronary circulation contains vascular niches where VPCs arenested together with ECs, SMCs, and adventitial fibroblasts thatfunction as supporting cells. Conversely, the myocardium possessesstem cell niches in which MPCs are clustered together with inter-stitial fibroblasts and myocytes that operate as supporting cells.

Transcriptional Profile. To establish whether VPCs and MPCsrepresent two distinct categories of progenitors rather than acommon pool of primitive cells at different stages of commitment,the identity of these PC classes was defined by analyzing theirtranscriptional profile. We have used quantitative RT-PCR andexamined a set of genes involved in self-renewal, multipotentiality,and lineage specification (Fig. S6 in SI Appendix). Several stemness-related genes were expressed in VPCs and MPCs. Among theself-renewal genes (5, 6), FGF4, telomerase, Myst1 and Myst2, andSox1 were represented equally in VPCs and MPCs. However,GDF3 was 3.5-fold higher in MPCs than in VPCs (Fig. 3). GDF3is part of a group of genes that flank Nanog, forming a chromatinloop of regulatory elements. The integrity of this region of thegenome is maintained by Oct4 and is essential for the preservationof self-renewal in embryonic stem cells (7). Genes that modulateasymmetric division (3) Numb, Pard6A, and Prox1 were uniformlypresent, ensuring the ability of VPCs and MPCs to self-renew, andcreate adequate progeny. Multiple cell cycle regulators, includingcyclins D1, D2, A2, and cdc2 and cdc42 showed comparable levelsof transcripts in VPCs and MPCs. These data indicate that both cellcategories possessed high intrinsic self-renewal ability and growthreserve typical of PCs (Fig. 3).

VPCs and MPCs showed nearly identical levels of Sox2, c-Myc,

Fig. 1. Localization and properties of vascular progenitor cells (VPCs) and myocyte progenitor cells (MPCs). (A) Cross-section of human epicardial coronary arterycomposed of several layers of smooth muscle cells (SMCs) [�-smooth-muscle-actin (�-SMA); red]. The c-kit-positive cells (green) are included in six rectangles. Three ofthe six rectangles are shown at a higher magnification in the adjacent panels. The c-kit-positive cells express KDR (white). Connexin 43 (Cx43; yellow dots; arrows) isseen between c-kit-KDR-positive cells and endothelial cells (von Willebrand factor; bright blue), SMCs (�-SMA; red), and adventitial fibroblasts (procollagen, procoll;magenta). The c-kit-KDR-positive cells are negative for CD45 and tryptase, excluding mast cells. Insets: Positive controls. See Fig. S1A in SI Appendix for the other threeareas. (B) FACSofVPCsandMPCs.Negativecontrolswereusedforallepitopes.Twoexamplescorrespondingto isotype-matchedantibodies forc-kitandKDRareshown.(C) Clones derived from single sorted VPCs and MPCs.

15886 � www.pnas.org�cgi�doi�10.1073�pnas.0907622106 Bearzi et al.

See Retraction Published September 30, 2019

Klf4, and Oct4 (Fig. 3), which are the four genes that promotereprogramming of fibroblasts into inducible pluripotent stem cells(8). This finding suggests that VPCs and MPCs are multipotent cellscharacterized by a significant degree of plasticity. However, vascu-lar-restricted genes and myocyte-specific genes appeared to bepoised for expression in VPCs and MPCs, respectively. The mRNAquantities of the EC transcription factor Vezf1, vascular adhesionprotein VCAM1, and the EC markers eNOS, vWf, and multimerinwere higher in VPCs than those in MPCs. Similarly, transcripts forGATA6 and contractile proteins �-SMA, SM22�, and smoothelinwere more expressed in VPCs than in MPCs. Conversely, MPCsshowed increased transcripts for Nkx2.5, �-myosin heavy chain, andmyosin heavy chain 7b (Fig. S7 in SI Appendix). Although thesedifferences in gene expression did not reach statistical significancedue to the undifferentiated state of the cells under our cultureconditions, VPCs and MPCs manifested a preferential lineagepotential; specific cell phenotypes became apparent when commit-ment was induced (Fig. 2A and Fig. S2 in SI Appendix).

Among the transcripts of the Notch pathway, the Delta-like 3ligand and the downstream regulator DTX1 were 2- and 3.5-foldmore abundant in MPCs than VPCs, respectively (Fig. 3). Impor-tantly, the Notch1 receptor is a critical determinant of the transitionof MPCs to amplifying myocytes (9). The endodermal transcriptionfactor Pdx1 was �4-fold higher in MPCs than in VPCs. Although

its role in MPCs remains to be determined, Pdx1 is locatedupstream of several members of the Nkx family. Two genes relevantto vascular cell turnover and repair, PPAR-� and Klf5, wereup-regulated, respectively, 10- and 2-fold in VPCs. PPAR-� ispresent in ECs and SMCs and is implicated in vessel homeostasis,whereas Klf5 is expressed abundantly in vascular structures duringdevelopment and in response to injury (10). These findings at thetranscriptional level, together with the data at the protein level atbaseline (Fig. 1B) and after differentiation (Fig. 2A and Fig. S2 inSI Appendix), are consistent with the notion that VPCs and MPCsare separate classes of PCs with distinct biological properties andspecific independent functions.

Conductive Coronary Arteries. Human MPCs can generate a largenumber of myocytes in the infarcted heart of immunodeficient miceor immunosuppressed rats (11). MPCs also form a small number ofresistance coronary arterioles and capillaries, which is consistentwith the multipotentiality of this PC class in vitro. The question wasthen whether human VPCs are able to create epicardial coronaryarteries in dogs with defects in coronary perfusion. A criticalcoronary stenosis was induced in the left anterior descendingcoronary artery (LAD) of immunosuppressed dogs. Subsequently,human VPCs, infected with EGFP, were injected above, lateral to,and below the site of constriction (Fig. S8 in SI Appendix). Dogs with

A

H

B

F

G

C

D E

Fig. 2. Differentiation and growth of vascular progenitor cells (VPCs) and myocyte progenitor cells (MPCs). (A) Percentage of VPCs and MPCs differentiating intoendothelial cells (ECs), smooth muscle cells (SMCs), and myocytes. *, P � 0.05 vs. ECs and SMCs. (B and C) Mitotic VPCs (arrows) show uniform (B) and nonuniform (C)distribution of �-adaptin (green), documenting symmetric and asymmetric division, respectively. (D–F) Daughter cells of symmetrically dividing VPCs (asterisks) do notexpressEts1 (D,blue),whereasonedaughtercellofasymmetricallydividingVPCs (arrows) showEts1 (E,blue;arrowhead)orGATA6(F, yellow;arrowhead). (G)Daughtercells of symmetrically dividing MPCs (asterisks) do not express GATA4 (white), whereas one daughter cell of asymmetrically dividing MPCs (arrows) shows GATA4(arrowhead). (H) Percentage of symmetrically (s) and asymmetrically (a) dividing VPCs and MPCs. *, P � 0.05 vs. symmetric division.

Bearzi et al. PNAS � September 15, 2009 � vol. 106 � no. 37 � 15887

MED

ICA

LSC

IEN

CES

See Retraction Published September 30, 2019

critical stenosis were selected because this model mimics thedevelopment of ischemic cardiomyopathy in humans. Additionally,dogs with critical stenosis cannot increase significantly coronaryblood flow (CBF) at rest or after stressful conditions. The gener-ation of collaterals never results in the restoration of CBF in thepresence of an occluded major coronary artery (12).

At 10 days after cell implantation, there were no changes in meanand phasic CBF, and the absence of reactive hyperemia wascomparable to baseline (Fig. 4A). However, at 30 days, a slow returnof CBF was detected after release of the transiently occluded LAD(Fig. 4B), suggesting the formation of coronary vessels in thepresence of critical stenosis (n � 9). These changes in CBF were notdictated by perfusion of the occluded LAD but by the vesselsgenerated around the affected artery. Histologically, large devel-oping coronary arteries of human origin were found in proximity tothe stenotic vessels (Fig. 4 C and D and Fig. S9 in SI Appendix). TheSMCs and ECs within the vessel wall were all positive for EGFP,and the human origin of these cells was confirmed by the detectionof human DNA sequences with an Alu probe.

At times, during coronary catheterization, the circumflex coro-nary artery (CX) and the stenotic LAD were seen clearly. AfterLAD occlusion, the CX remained visible, and the distal portion ofthe LAD was no longer apparent (Fig. 4E). However, left ventric-ular pressure, dP/dt, and segment length function in the ischemicand nonischemic regions of the wall did not change (Fig. 4F),suggesting that CBF was not compromised further by LAD occlu-sion. Gold-labeled microspheres then were injected with the LADopen to establish baseline CBF, whereas lutetium-labeled micro-

Fig. 3. Transcriptional profile. Three-dimensional representation of the tran-scriptionalprofileofvascularprogenitor cells (VPCs)andmyocyteprogenitor cells(MPCs). Bar graphs illustrate the expression of genes discussed in the text. *, P �0.05 between VPCs and MPCs.

Fig. 4. Generation of large coronary arteries. (A and B) Dogs with coronary stenosis injected with EGFP-labeled human vascular progenitor cells (VPCs). (A) At 10 days,the lack of reactive hyperemia after release of a 15-s occlusion of the left anterior descending coronary artery (LAD) is shown by both mean and phasic coronary bloodflow (CBF). (B) At 30 days, there was a slow return of CBF (arrows) after release of LAD occlusion. The transient occlusion was identical in A and B, although the returnofflowwasfollowedfora longerperiodinB. (CandD)Arteries0.9and1.3mmindiameterweredetected.All smoothmusclecellsarepositivefor�-smooth-muscle-actin(�-SMA) (red), EGFP (green), �-SMA/EGFP (yellowish), and Alu probe (white). Preexisting coronary vessels (C, arrows) are �-SMA positive and EGFP and Alu probenegative. (E) Coronary angiogram in a treated dog at 47 days shows the stenotic LAD and the normally perfused coronary artery (CX) before (left) and after (right) LADocclusion. (F) Left ventricular pressure, dP/dt, and segment length function in the ischemic and nonischemic regions of the ventricular wall before (left) and after (right)LAD occlusion. Complete LAD occlusion did not affect left ventricular pressure, dP/dt, and segment length function (compare left and right).

15888 � www.pnas.org�cgi�doi�10.1073�pnas.0907622106 Bearzi et al.

See Retraction Published September 30, 2019

spheres were administered during LAD occlusion. Sections werecollected from the region proximal to the region distal to theocclusion to evaluate the distribution of CBF, whereas every othersection was used for histology (Fig. S10 in SI Appendix).

In each section, CBF at baseline was �1 mL/min/g. With LADocclusion, CBF increased to �2 mL/min/g (Fig. 5A). Thus, CBF

increased nearly 2-fold in the fully dilated coronary circulationdistal to the stenotic vessel (n � 8). Histologically, large andintermediate newly formed coronary arteries together with resis-tance arterioles and capillary structures were identified (Fig. 5B andFigs. S11 and S12 in SI Appendix). Together, these observationsstrongly suggest that the regenerated large conductive coronaryarteries were connected with the LAD above the site of constric-tion, as previously demonstrated in the rat model of coronaryocclusion (13), creating a biological bypass. Although hearts werefixed by perfusion, residual red blood cells were detected in thevarious vessel categories, documenting their functional competenceand integration with the primary coronary circulation (Fig. S12 inSI Appendix). Human ECs and SMCs in the vessel wall had 2n DNAcontent and possessed at most one human X-chromosome andnever dog X-chromosome (Fig. S13 in SI Appendix), excludingfusion events. Vasculogenesis involved all components of thecoronary circulation from vessels �100 �m to 1.5 mm in diameter(Fig. 5C), pointing to vessel regeneration as the mechanism ofenhanced CBF and tissue oxygenation in the potentially ischemicmyocardium. The increase in CBF is proportional to the fourthpower of the increase in the aggregate diameter of the coronarycirculation CBF may be doubled by a 16% increase in the numberof vessels. The improvement in CBF was coupled with increasedsegment length function in the ischemic myocardium (Fig. 5D). Athin layer of human cardiomyocytes was detected in the epicardialsurface (Fig. S14 in SI Appendix).

DiscussionThe results of the current study indicate that the human heartpossesses two classes of PCs that have powerful but distinctvasculogenic and myogenic properties. The VPCs are nested invascular niches distributed throughout the coronary circulation,and MPCs are clustered in stem cell niches located in themyocardial interstitium. The ECs, SMCs, and adventitial fibro-blasts are responsible for the anchorage of VPCs to the nichestructure, whereas myocytes and fibroblasts function as support-ing cells for MPCs in myocardial niches. The VPCs differentiatepredominantly into ECs and SMCs, and MPCs acquire preva-lently the cardiomyocyte phenotype. Potentially, the humanheart has the inherent ability to restore the various segments ofthe coronary circulation by activation and commitment of VPCstogether with the opportunity to regenerate losses of musclemass by growth and differentiation of MPCs. Various propor-tions of VPCs and MPCs may be used according to the needs ofthe organ and the type of the cardiac lesion.

Analysis of the transcriptome allows the recognition of genesexpressed differentially in stem cell classes and helps to identifydivergent patterns and functional background. Transcriptionalprofiling was used here to define the molecular signatures ofVPCs and MPCs. The self-renewal property of stem cells isregulated by a multitude of genes, including FGF4 and telom-erase (5, 6). TERT and FGF4 did not differ in VPCs and MPCs,but whether this molecular characteristic is sufficient to controlthe kinetics of these PC classes, favoring in both cases thegeneration of daughter stem cells could not be established. Sox2,c-Myc, Klf4, and Oct4 reprogram the growth and differentiationbehavior of adult somatic cells and govern the proliferation andpluripotency of embryonic stem cells (7, 8). Their expression wascomparable in VPCs and MPCs, suggesting that cardiac PCspossess similar levels of plasticity and that stemness and multi-potentiality may be regulated by the same group of genes.However, the differential expression of the components of theNotch pathway in VPCs and MPCs suggests that the fate of eachPC pool is defined partly by a distinct set of genes. Activation ofNotch and Wnt modulates cardiomyogenesis in the embryonicand postnatal heart (9, 14). Some of the genes involved in lineagespecification of cardiac PCs into myocytes and vascular cells wereexpressed preferentially in MPCs and VPCs, respectively. This

Fig. 5. CBF and newly formed vasculature. (A) Distribution of CBF at baseline (b,green bars) and after left anterior descending coronary artery occlusion (o, bluebars) in sections proximal and distal to the stenosis. For details, see text and Fig.S10 in SI Appendix. *, P � 0.05 vs. baseline; n � 8. (B) Two newly formed coronaryarteries from two hearts in which CBF was measured. Each vessel is positive for�-SMA (red) and Alu (white). The diameter of each vessel is indicated. Theexpression of EGFP alone and with �-SMA for each vessel is shown in Fig. S11 inSI Appendix. (C) Classes of newly formed human coronary vessels within the dogheart. (D)Tracingof segment length inthe ischemicmyocardiumat1and4weeksin thesamedog. (Lower) Improvement in shorteningof the ischemicmyocardiumwith time. *, P � 0.05 vs. 1 week.

Bearzi et al. PNAS � September 15, 2009 � vol. 106 � no. 37 � 15889

MED

ICA

LSC

IEN

CES

See Retraction Published September 30, 2019

may reflect the restriction in developmental options that occursin tissue-specific adult PCs.

A critical question concerns whether these PC categoriesoriginate, live, and die within the heart or whether the bonemarrow replenishes the coronary vasculature and the myocar-dium with undifferentiated cells. In vascular and myocardialniches, the pool size of VPCs and MPCs may be preserved bymigration of PCs from the bone marrow, asymmetric division ofresident PCs within vascular and myocardial niches, or both. Thedocumentation that VPCs and MPCs divide symmetrically andasymmetrically in vitro and in vivo (11) suggests that the need forthe contribution of bone marrow PCs to cardiac homeostasis inhumans may not be essential. The involvement of the bonemarrow in human cardiac chimerism has been claimed (15), butthe degree of chimerism in cardiac allograft and in the hearts ofpatients who received allogeneic bone marrow transplantationdiffers significantly, being much higher in the transplanted heart(16). With parabiosis, in the absence of myocardial damage,circulating bone marrow PCs do not home to the heart (17).Collectively, these observations indicate that in the human heartthe PC compartment is regulated largely by intrinsic cellularmechanisms, although extrinsic cellular processes may partici-pate in cardiac repair after injury.

The treatment of coronary artery disease has improved dra-matically in recent years. However, morbidity and mortality forischemic cardiomyopathy continue to increase and parallel theextension in lifespan of the population, pointing to aging as themajor risk factor of human heart failure. Bypass surgery andcatheter-based reinstitution of CBF have been introduced suc-cessfully, but these interventions correct only in part the vasculardefects and are limited by the number of surgical grafts, thepossibility of restenosis, and the complexity of reintervention(18). These variables underscore the need for the developmentof new strategies for the management of coronary atheroscle-rosis in humans.

In the current study we have identified and characterized aresident human VPC that can form large conductive coronaryarteries and their distal branches, correcting, at least in part,alterations in blood flow created by prolonged coronary con-striction in chronically instrumented conscious dogs. Cell ther-apy with the generation of coronary vessels by human bonemarrow PCs has been obtained previously, but this has beenrestricted to capillary profiles and occasionally small resistancearterioles, which enhance tissue oxygenation but have littleimpact on flow regulation. Restriction in myocardial perfusionmediated by stenosis of a major epicardial coronary artery wastreated by the delivery of human VPCs, which differentiated intoECs and SMCs organized in coronary vessels up to 1.5 mm inluminal diameter. The VPCs did not integrate in preexistingnonfunctional collateral vessels, contributing to their maturationin working vascular structures. The newly formed conductivearteries were made exclusively of human SMCs and ECs, whichwere the progeny of the injected PCs. The human origin of theregenerated vessels was also apparent in the other segments ofthe newly formed coronary vasculature. From a clinical perspec-tive, regeneration of large conductive arteries may require thelocal administration of VPCs, whereas intracoronary and in-tramyocardial injection of MPCs may be equally effective inpromoting myocyte regeneration. Caution has to be exercised inthe interpretation of our data, because the long-term outcomeof this type of cell therapy remains unknown.

Materials and MethodsMyocardial specimens from 12 patients were studied. Two classes of PCs wereisolated, characterized, and studied in vitro and in vivo. See SI Materials andMethods in SI Appendix.

ACKNOWLEDGMENTS. S.B. was supported by a grant from CardiocentroTicino, Lugano, Switzerland.

1. Abdel-LatifA,etal. (2007)Adultbonemarrow-derivedcells forcardiac repair:Asystematicreview and meta-analysis. Arch Intern Med 167:989–997.

2. OlsonEN(2006)Generegulatorynetworks in theevolutionanddevelopmentof theheart.Science 313:1922–1927.

3. Urbanek K, et al. (2006) Stem cell niches in the adult mouse heart. Proc Natl Acad Sci USA103:9226–9231.

4. Yang L, et al. (2008) Human cardiovascular progenitor cells develop from a KDR� embry-onic-stem-cell-derived population. Nature 453:524–528.

5. Morrison SJ, Prowse KR, Ho P, Weissman IL (1996) Telomerase activity in hematopoieticcells is associated with self-renewal potential. Immunity 5:207–216.

6. Mayshar Y, et al. (2008) Fibroblast growth factor 4 and its novel splice isoform haveopposing effects on the maintenance of human embryonic stem cell self-renewal. StemCells 26:767–774.

7. Levasseur DN, Wang J, Dorschner MO, Stamatoyannopoulos JA, Orkin SH (2008) Oct4dependenceofchromatinstructurewithintheextendedNanoglocus inEScells.GenesDev22:575–580.

8. Gurdon JB, Melton DA (2008) Nuclear reprogramming in cells. Science 322:1811–1815.9. Boni A, et al. (2008) Notch1 regulates the fate of cardiac progenitor cells. Proc Natl Acad

Sci USA 105:15529–15534.

10. Nagai R, Suzuki T, Aizawa K, Shindo T, Manabe I (2005) Significance of the transcriptionfactor KLF5 in cardiovascular remodeling. J Thromb Haemost 3:1569–1576.

11. Bearzi C, et al. (2007) Human cardiac stem cells. Proc Natl Acad Sci USA 104:14068–14073.12. Lavallee M, Cox DA, Patrick TA, Vatner SF (1983) Salvage of myocardial function by

coronary artery reperfusion 1, 2, and 3 hours after occlusion in conscious dogs. Circ Res53:235–247.

13. Tillmanns J, et al. (2008) Formation of large coronary arteries by cardiac progenitor cells.Proc Natl Acad Sci USA 105:1668–1673.

14. Kwon C, et al. (2007) Canonical Wnt signaling is a positive regulator of mammalian cardiacprogenitors. Proc Natl Acad Sci USA 104:10894–10899.

15. Schwartz RS, Curfman GD (2002) Can the heart repair itself? N Engl J Med 346:2–4.16. Thiele J, et al. (2004) Mixed chimerism of cardiomyocytes and vessels after allogeneic bone

marrow and stem-cell transplantation in comparison with cardiac allografts. Transplan-tation 77:1902–1905.

17. Wagers AJ, Sherwood RI, Christensen JL, Weissman IL (2002) Little evidence for develop-mental plasticity of adult hematopoietic stem cells. Science 297:2256–2259.

18. Flaherty JD, Davidson CJ (2005) Diabetes and coronary revascularization. J Am Med Assoc293:1501–1508.

15890 � www.pnas.org�cgi�doi�10.1073�pnas.0907622106 Bearzi et al.

See Retraction Published September 30, 2019