ARTICLEdoi:10.1038/nature11139

Heart repair by reprogrammingnon-myocytes with cardiac transcriptionfactorsKunhua Song1, Young-Jae Nam1,2, Xiang Luo2, Xiaoxia Qi1, Wei Tan2, Guo N. Huang1, Asha Acharya1, Christopher L. Smith1,Michelle D. Tallquist1, Eric G. Neilson3, Joseph A. Hill1,2, Rhonda Bassel-Duby1 & Eric N. Olson1

The adult mammalian heart possesses little regenerative potential following injury. Fibrosis due to activation of cardiacfibroblasts impedes cardiac regeneration and contributes to loss of contractile function, pathological remodelling andsusceptibility to arrhythmias. Cardiac fibroblasts account for a majority of cells in the heart and represent a potentialcellular source for restoration of cardiac function following injury through phenotypic reprogramming to a myocardialcell fate. Here we show that four transcription factors, GATA4, HAND2, MEF2C and TBX5, can cooperatively reprogramadult mouse tail-tip and cardiac fibroblasts into beating cardiac-like myocytes in vitro. Forced expression of these factorsin dividing non-cardiomyocytes in mice reprograms these cells into functional cardiac-like myocytes, improves cardiacfunction and reduces adverse ventricular remodelling following myocardial infarction. Our results suggest a strategy forcardiac repair through reprogramming fibroblasts resident in the heart with cardiogenic transcription factors or othermolecules.

Heart disease caused by the loss or dysfunction of cardiomyocytes is theleading cause of death worldwide1. Whereas the neonatal mammalianheart can regenerate following injury2, the capacity for regeneration ofadult mammalian hearts is limited3. A fundamental goal of regenerativecardiovascular medicine is to successfully repair injured hearts byreplacing cardiomyocytes and decreasing fibrosis. Transplantation ofcardiac stem cells or stem cell-derived cardiomyocytes to improvecardiac function holds clinical potential, but is relatively inefficient4–6.

Fibroblasts can be reprogrammed into pluripotent stem cells,muscle cells and neurons by combinations of lineage-enriched tran-scription factors7–12. The human heart is composed of approximately30% cardiomyocytes and 60–70% cardiac fibroblasts 13. We sought todefine the optimal combination of core cardiac transcription factorsnecessary and sufficient for reprogramming of adult fibroblasts intofunctional cardiomyocytes and to determine if these factors couldimprove cardiac function in injured hearts through reprogrammingof endogenous cardiac fibroblasts. Here we show that four transcriptionfactors, GATA4, HAND2, MEF2C and TBX5, are able to reprogramadult mouse fibroblasts into functional cardiac-like myocytes in vitroand in vivo, and expression of these factors in non-cardiomyocytesenhances function of injured hearts following myocardial infarction.

Reprogramming of adult fibroblasts into cardiac-likemyocytesA core set of evolutionarily conserved transcription factors (GATA4,HAND2, MEF2C, MESP1, NKX2-5 and TBX5) controls cardiac geneexpression and heart development14,15. Recently, GATA4, MEF2Cand TBX5 were reported to be capable of converting neonatalfibroblasts to cardiomyocyte-like cells in vitro16. To search for anoptimal combination of transcription factors capable of inducingcardiac lineage-reprogramming, we generated retroviruses to expresseach of the six core cardiac transcription factors (GATA4 (G),

HAND2 (H), MEF2C (M), MESP1 (Ms), NKX2-5 (N), and TBX5(T)) in fibroblasts derived from adult mice bearing a cardiac-specifica-MHC-GFP transgene, where the promoter of the a-myosin heavychain gene controls the expression of green fluorescent protein(Supplementary Figs 1 and 2).

Potential cardiogenic activity of the six factors (G, H, M, Ms, N, T)in adult tail-tip fibroblasts (TTFs) from a-MHC-GFP transgenic micewas quantified by analysis of GFP1 cells by flow cytometry 9 daysafter viral transduction. No GFP1 cells were observed in fibroblastcultures transduced with viral backbone or without infection.However, the six factors together generated a small population of cells(,7%) positive for GFP (Supplementary Fig. 3). After several roundsof withdrawing one factor, we identified several combinations, includ-ing GHMMsT, GHMT, HMMsT, GMT, HMT and MT, that werecapable of efficiently inducing a-MHC-GFP1 cells (SupplementaryFig. 3). The cooperativity amongst the different factors is consistentwith their ability to synergistically activate cardiac gene expressionand to activate each other’s expression14,17–20.

To determine whether the above factors could activate endogenouscardiac-specific genes in adult TTFs, we examined expression ofcardiac troponin T (cTnT, also known as TNNT2) and a-MHC-GFP by flow cytometry (Supplementary Fig. 4). Depending on theexperiment, GHMT induced between 5.2% and 19.7% (average 9.2%)of cells to become positive for both a-MHC-GFP and cTnT. We referto GHMT-transduced cells expressing cardiomyocyte markers asinduced cardiac-like myocytes (iCLMs). By comparison, GMTinduced approximately 2.9% (1.5% to 5.6%) of adult TTFs to adopta cTnT1 a-MHC-GFP1 phenotype (Supplementary Fig. 4). Thepercentage of iCLMs generated from adult TTFs by GHMT (9.2%)was about fourfold higher than previously reported for neonatalTTFs with GMT (2.5%)16. The percentage of cells expressingcTnT and a-MHC-GFP reached a peak at day 7 of transduction;

1Department of Molecular Biology, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, Texas 75390-9148, USA. 2Department of Internal Medicine, University of TexasSouthwestern Medical Center, 5323 Harry Hines Blvd, Dallas, Texas 75390-9148, USA. 3Department of Medicine, Vanderbilt University School of Medicine, Nashville, Tennessee 37232, USA.

3 1 M A Y 2 0 1 2 | V O L 4 8 5 | N A T U R E | 5 9 9

Macmillan Publishers Limited. All rights reserved©2012

the fraction of iCLMs declined thereafter due to overgrowth of non-reprogrammed fibroblasts (Supplementary Fig. 5).

Cardiac fibroblasts are the most prevalent interstitial cell type inadult mammalian hearts. We transduced adult cardiac fibroblastswith GHMT, GMT or empty viruses and analysed expression ofcardiac markers by flow cytometry 1 week later. GHMT induced6.8% of cardiac fibroblasts to become cTnT1 a-MHC-GFP1,compared with 1.4% double-positive cells with GMT (Supplemen-tary Fig. 6a). GHMT induced both cTnT and cTnI (cardiac troponinI, also known as TNNI3) in 7.5% of cells (Supplementary Fig. 6b).Thus, GHMT represented the most optimal combination of factorsfor efficient initiation of cardiac gene expression in adult fibroblasts.a-MHC-GFP1 cells derived from adult TTFs and cardiac fibroblastsby GHMT transduction showed strong immunostaining of thesarcomeric proteins a-actinin and cTnT (Fig. 1a, b and Supplemen-tary Fig. 7). More organized sarcomeres were observed in iCLMs atday 30 (Fig. 1b) compared to day 14 (Fig. 1a). In the presence ofGHMT, three types of iCLMs, referred to as types A, B and C, whichseem to represent a spectrum of cardiac reprogramming, wereinduced from adult cardiac fibroblasts and TTFs. Type A iCLMs onlyexpressed a-MHC-GFP. Type B iCLMs expressed both a-MHC-GFPand a-actinin. Type C iCLMs not only expressed a-MHC-GFP anda-actinin, but also showed sarcomere-like structures (Fig. 1a). GHMTinduced ,15% of adult cardiac fibroblasts and ,18% of adult TTFs tobecome a-MHC-GFP-positive cells, and ,30% and ,10% of thesebecame type C iCLMs, respectively (Fig. 1c).

Microarray and quantitative PCR (qPCR) analysis of gene expres-sion patterns showed expression of a broad range of cardiac genes,indicative of a cardiac-like phenotype, and concomitant suppressionof non-myocyte genes including Fsp1 (fibroblast-specific protein 1 orS100A4) in fibroblasts transduced with GHMT (Supplementary Figs 8and 9). Following maintenance of GHMT-transduced adult cardiacfibroblasts or TTFs in culture for more than 5 weeks, we observedspontaneous contractions in approximately 0.2% of type C iCLMs.These beating iCLMs had calcium transients and action potentials(Supplementary Movies 1, 2 and 3, Fig. 1d and Supplementary Figs10 and 11). iCLMs had a pattern of calcium transients most similar toneonatal ventricular cardiomyocytes (Fig. 1d). Transduction withinducible expression vectors showed that the cardiomyocyte-likephenotype was stable following termination of exogenous GHMTexpression (Supplementary Fig. 12). Co-staining for cardiomyocytemarker and Myc-tagged GHMT indicates that induction of thecardiomyocyte-like phenotype by GHMT is cell-autonomous (Sup-plementary Fig. 13). Together, these results indicate that GHMT canactivate cardiac gene expression in a sub-population of TTFs andcardiac fibroblasts. The relatively small percentage of cells that adoptsthe cardiac-like phenotype perhaps indicates a precise stoichiometryor high level of the cardiac factors required for phenotypic conver-sion, which is achieved in only a fraction of cells, or to differentialsusceptibility to reprogramming amongst heterogeneous cell typesin the starting non-myocyte population. The efficiency of cellularreprogramming to iCLMs by GHMT is comparable to that of repro-gramming of induced pluripotent stem (iPS) cells by pluripotencyfactors7,21.

Reprogramming of FSP1-tagged cells into iCLMs in vivoFollowing myocardial infarction or other forms of cardiac injury,cardiomyocytes undergo necrotic and apoptotic cell death and cardiacfibroblasts are activated to produce collagen and other extracellularmatrix components, causing fibrosis and impaired cardiac function22.Therefore, we sought to determine whether reprogramming cardiacfibroblasts to a cardiomyocyte fate might blunt the decline in cardiacfunction post-myocardial infarction. Adult mammalian cardiomyocytesdo not divide and are therefore resistant to retroviral expression. Weexpressed the cardiogenic transcription factors in cardiac fibroblasts andother dividing cells in vivo using a retrovirus expression system, which is

specific for proliferating cells and directs expression preferentially inmyofibroblasts of injured rodent hearts23. We confirmed the specificityof retroviral infection for replicating non-cardiomyocytes by injectingconcentrated GFP retroviruses into injured hearts after ligation ofthe left anterior descending coronary artery (LAD), which induces

a

Typ

e A

iC

LM

s

α-M

HC

-GF

P+

Typ

e B

iC

LM

s

α-M

HC

-GF

P+,

α-a

ctin

in+

Typ

e C

iC

LM

s

α-M

HC

-GF

P+,

α-a

ctin

in+,

‘sarc

om

ere

s’

α-actinin, DAPIGFP, DAPI Merge

b c

Fra

ctio

n o

f

α-M

HC

-GF

P+ c

ells

(%

)

A B

CF

C0

Cell type:

10

30

20

50

70

40

60

A B C

TTF

CF TTF

α-M

HC

-GF

P+ c

ells

(%

)

Vecto

r

GH

MT

Vecto

r

GH

MT0

5

10

15

30

20

25F

ura

-2 r

atio

(340/3

80 n

m)d iCLM

Time (seconds)200 40

1.0

0.8

1.2Neonatal ventricular

cardiomyocyte

Time (seconds)200 40

1.0

0.8

1.2Adult cardiac

fibroblast

Time (seconds)200 40

1.0

0.8

1.2

α-actinin, DAPI

Figure 1 | Reprogramming fibroblasts towards a cardiac phenotype in vitroby GHMT. a, b, Immunofluorescent staining for cardiac markers, a-MHC-GFP and a-actinin. Adult cardiac fibroblasts (CF) isolated from a-MHC-GFPreporter mice were transduced by retroviruses carrying GHMT or empty vectoralone. Immunocytochemistry was performed at 14 days (a) or 30 days(b) following transduction. Scale bars, 20mm. Three types of iCLMs werecategorized on the basis of cardiac gene expression and morphology. Type AiCLMs only express a-MHC-GFP (top panels). Type B iCLMs express botha-MHC-GFP and a-actinin, but do not have sarcomeric structures (middlepanels). Type C iCLMs express both a-MHC-GFP and a-actinin, and havesarcomeric structures (bottom panels). Sarcomeres were more organized at day30 than at day 14 post-transduction. White boxes are enlarged in insets.c, Quantification of type A, B and C iCLMs in adult cardiac fibroblasts andTTFs 14 days after transduction with GHMT. Ten fields were randomly chosenfrom each experiment. Three independent experiments were performed. Dataare presented as mean 6 s.d. d, Representative calcium transient traces fromthe indicated cell types depicted as Fura-2 ratios (340/380 nm). iCLMs werederived from adult cardiac fibroblasts and showed a pattern of calciumtransients most similar to neonatal ventricular cardiomyocytes. Adult cardiacfibroblasts do not display calcium transients.

RESEARCH ARTICLE

6 0 0 | N A T U R E | V O L 4 8 5 | 3 1 M A Y 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

myocardial infarction. GFP expression was clearly detected in theischaemic area. However, none of the GFP1 cells expressed the cardiacmarker, cTnT, consistent with the specificity of retroviral infection forproliferating non-cardiomyocytes (Supplementary Fig. 14).

FSP1 is expressed in non-cardiomyocytes such as fibroblasts andtransitioning epithelia24–26. In mouse and human hearts, expression ofFSP1 primarily colocalizes with markers of cardiac fibroblasts andincreases following myocardial infarction26. Non-cardiomyocytes inmice carrying alleles of Fsp1-Cre and Rosa26-LacZ are specificallylabelled with b-galactosidase (b-gal), providing a marker for fibroblastlineage tracing24. To examine whether cardiac transcription factorswere able to activate cardiac genes in non-cardiomyocytes in vivo,we performed LAD ligation on Fsp1-Cre/Rosa26-LacZ mice andinjected concentrated retroviruses encoding GHMT or GFP into theborder zone immediately following LAD ligation. We then analysedb-gal activity in histological sections of hearts at various times thereafter.In uninjured hearts, less than one b-gal1 cardiomyocyte per sectionwas observed. After injury, b-gal expression was readily detected incardiac fibroblasts throughout the infarct zone (Fig. 2a, b). In injuredhearts infected with GFP viruses, 0.05 6 0.13% of cardiomyocytes inthe injured area were b-gal1 (Fig. 2c), which may be due to low-levelectopic activation or to a basal level of new cardiomyocyte formationfrom a stem cell pool, as reported previously27. In contrast, abundantclusters of intensely stained b-gal1 cardiomyocytes were observedthroughout the infarct and border zone of injured hearts infected withthe GHMT retrovirus cocktail (Fig. 2b). We observed that 6.5 6 1.2%of cardiomyocytes in the injured area showed b-gal activity (Fig. 2c).Generally, more b-gal1 cardiomyocytes were observed in the borderzone adjacent to the infarct region, which may be due to intact vascularstructures or higher viral infection in this region. Similar results wereobtained upon injection with GHMMsT, whereas the inclusion ofNKX2-5 (GHMMsNT) diminished the efficacy of the other five factors(Supplementary Fig. 15), as seen in vitro. b-gal1 cardiomyocytesexpressed cTnT and showed clear striations 3 weeks after viral trans-duction (Fig. 2b). Through quantification of b-gal1 cardiomyocytes inhistological sections of infarcted hearts, we calculated that at least10,000 new myocytes were generated in the injured area after 3 weeksof GHMT infection. This is likely to be an underestimate of the numberof iCLMs generated in vivo because the reporter is only expressed in asubset of non-myocytes, such that unmarked cells could also be repro-grammed but go undetected.

In the heart, gap junctions composed of connexins ensure electricaland metabolic coupling between cardiomyocytes and coordinate theircontractility28. To determine whether reprogrammed cardiomyocytescould couple with surrounding endogenous myocytes through gap junc-tions, we performed immunostaining for connexin 43 (Cx43, also knownas GJA1), the major connexin in functional cardiomyocytes29. Gap junc-tions were observed between b-gal1 and b-gal– cardiomyocytes andbetween b-gal1 cardiomyocytes (Fig. 2d), indicating coupling of repro-grammed cardiomyocytes with surrounding myocytes. Reprogrammedb-gal1 cardiomyocytes isolated from ventricular myocardium andidentified by labelling with a fluorogenic, lipophilic b-gal substrate(C12FDG) also had a pattern of contractility and calcium transientssimilar to normal b-gal– ventricular cardiomyocytes in response to elec-trical pacing at 1 Hz (Fig. 2e), indicating their functionality.

Reprogramming of TCF21-tagged cells into iCLMs in vivoTo rule out the possibility that injury or viral infection might somehowactivate the Fsp1 reporter in cardiomyocytes, we generated a strain ofmice harbouring an inducible MerCreMer expression cassette insertedby homologous recombination into the Tcf21 (also known as capsulinor epicardin) locus (Tcf21iCre), which is expressed specifically in non-cardiomyocytes in the heart (Supplementary Fig. 16)30. Intercrossing ofthese mice with mice bearing the Cre-inducible R26RtdTomato reportershowed specific expression of Tomato predominantly in cardiac fibro-blasts with less expression in endothelial cells (Supplementary Fig. 16).

[Ca

2+] i

(μM

)1 s

0.1

0.45

1 s0

0.15

Cell

sho

rtenin

g

(μm

)[C

a2+] i

(μM

)

1 s0.1

0.45

1 s0

0.15

Cell

sho

rtenin

g

(μm

)

1 s0

0.03

Cell

sho

rtenin

g

(μm

)

e β-gal+ cardiomyocyte (iCLM)β-gal– cardiomyocyte

Contraction

Ca2+ transient

Contraction

Contraction

Ca2+ transient

Matu

re iC

LM

Imm

atu

re iC

LM

a Uninjured Injured + GFPInjured + GFP Injured + GHMT

LV LV LV LV

b Uninjured

LacZ

cT

nT

1 2

3

4

Injured + GFP Injured + GHMT

1 2

3

4

c

β-g

al+

card

iom

yo

cyte

s

in inju

red

are

a (%

)

012

GFP

MI

GHMT

3456789

d LacZ

A

BD

C

cTnT Cx43

A

BD

C

Figure 2 | Reprogramming non-cardiomyocytes towards a cardiac fate invivo by GHMT. a, Heart sections of Fsp1-cre/Rosa26-LacZ mice uninjured or3 weeks post-myocardial infarction followed by injection of GFP or GHMTretroviruses. GFP-infected myocardium showed only b-galactosidase-positive(b-gal1) non-cardiomyocytes, GHMT-infected myocardium showed extensiveb-gal1 non-cardiomyocytes and cardiomyocytes. Black boxes in top panels areenlarged in lower panels. LV, left ventricle. Scale bars, 40mm. b, cTnTimmunostaining and X-gal staining of heart sections of Fsp1-cre/Rosa26-LacZmice uninjured or 3 weeks post-myocardial infarction followed by injection ofGFP or GHMT retroviruses. The same cell is marked with the same number inthe top and bottom panels (nos 1–4). Several b-gal1 cells in GHMT-infectedinjured hearts expressed cTnT and had organized sarcomere structure. Whiteboxes are enlarged in insets. Sections of injured hearts were taken at the borderzone. Scale bars, 40mm. c, Quantification of b-gal1 cardiomyocytes in borderzones and infarct zones from GFP-infected (168 sections, n 5 3) and GHMT-infected hearts of Fsp1-cre/Rosa26-LacZ mice (20 sections, n 5 2) post-myocardial infarction. Sections were taken at four levels with an interval of250mm below LAD ligation site. MI, myocardial infarction. Data are presentedas mean 6 s.d. d, Staining of border zone from GHMT-infected hearts of Fsp1-cre/Rosa26-LacZ mice post-myocardial infarction by X-gal (blue), anti-cTnT(red) and anti-Cx43 (green). Gap junctions (green) were observed betweenb-gal1 and b-gal2 cardiomyocytes (A and B) and between b-gal1

cardiomyocytes (C and D). The same cell is marked with the same letter. Scalebar, 20mm. e, Contractility and Ca21 transients of b-gal1 and b-gal2

cardiomyocytes. b-gal1 cardiomyocytes (iCLMs) were labelled with thefluorogenic b-gal substrate C12FDG (green). Traces of sarcomere shorteningwere recorded from field-stimulated b-gal2 cardiomyocytes (n 5 15) andiCLMs (n 5 7) with clear striated morphology. 71.4% of iCLMs had a similarpattern of contractility and Ca21 transients tob-gal2 cardiomyocytes. 28.6% ofiCLMs demonstrated immature contractility.

ARTICLE RESEARCH

3 1 M A Y 2 0 1 2 | V O L 4 8 5 | N A T U R E | 6 0 1

Macmillan Publishers Limited. All rights reserved©2012

However, we only observed one Tomato1 cardiomyocyte in over 40histological sections of normal hearts. Tcf21iCre R26RtdT mice weretreated with tamoxifen for 3 days to mark TCF21-expressing cells.Eight days after the last tamoxifen treatment, LAD ligation was per-formed and animals were injected with empty vector, GFP or GHMTretroviruses and analysed 3–4 weeks later. Numerous Tomato1

cardiomyocytes were observed in GHMT-injected hearts comparedto GFP-injected ones (Fig. 3a).

To quantify Tomato1 cardiomyocytes, we isolated cardiomyocytesfrom intact hearts using the Langendorff perfusion method. InGHMT-injected hearts, 2.4 6 1.5% of cardiomyocytes in the injuredarea were Tomato1, and these Tomato1 cardiomyocytes showedclear striations (Fig. 3b, c). However, in empty vector-injected injuredhearts, only 0.004 6 0.005% of cardiomyocytes were Tomato1

(Fig. 3c). Tcf21iCre marks a smaller, more restricted fraction of non-myocytes than Fsp1-Cre. Hence the number of new myocytes markedwith this reporter is lower than with Fsp1-Cre. These findings indicatethat forced expression of GHMT in non-myocytes of the heart issufficient to induce the expression of cardiac markers. The Tomato1

cardiomyocytes induced by GHMT in vivo were able to spontaneouslycontract in vitro (Supplementary Fig. 17 and Supplementary Movie 4).Recording of action potentials of reprogrammed Tomato1 cardiomyo-cytes and endogenous cardiomyocytes by whole-cell patch clampingindicated functionality of cardiomyocytes induced by GHMT in vivo(Fig. 3d).

To further test whether GHMT indeed promoted the formation ofnew cardiomyocytes following myocardial infarction, and to rule outthe possibility that labelled cardiomyocytes obtained from Fsp1-Creor Tcf21iCre/Rosa marker mice injected with GHMT might arise fromfusion of native cardiomyocytes with non-myocytes, we used micewith an inducible a-MHC-MerCreMer transgene and Rosa26-LacZreporter. Gavage of these mice with tamoxifen for seven consecutivedays resulted in labelling of 87.7 6 1.9% of cardiomyocytes in the leftventricular myocardium. Following LAD ligation and injection ofGFP retroviruses, we observed a reduction in the percentage ofb-gal-labelled cardiomyocytes (83.2 6 4.6%) in the border zoneadjacent to the infarct, indicating replenishment of cardiomyocytesfollowing injury, as described previously27,31. Injection of GHMTretroviruses further reduced the percentage of b-gal1 cardiomyocytesin the border zone (76.0 6 5.2%) (Supplementary Fig. 18). Thesefindings indicate that GHMT promotes the formation of newcardiomyocytes from a non-a-MHC lineage in vivo following injury.

Enhancement of cardiac regeneration of injured heartsWe examined whether forced expression of GHMT in non-cardiomyocytes could lead to measurable improvement in functionof ischaemic hearts. Cardiac function following myocardial infarctionwas assessed in a blinded fashion by fractional shortening , ejectionfraction and stroke volume using echocardiography and magneticresonance imaging (MRI). Twenty-four hours after LAD ligation,fractional shortening and ejection fraction assessed by echocardiographyof all mice decreased relative to sham-operated mice. Thereafter,cardiac function of GFP-injected mice continued to decline, reachinga stable value 2 weeks post-myocardial infarction; fractional shorten-ing < 13% and ejection fraction < 30%. In contrast, infection ofinjured myocardium with GHMT retroviruses blunted furtherworsening of cardiac function 3 weeks post-myocardial infarction;fractional shortening < 26% and ejection fraction < 51% (Fig. 4a).By comparison, functional improvement was delayed and less com-plete with GMT, consistent with the reduced efficiency of this tran-scription factor combination in reprogramming in vitro.

To determine whether functional improvement was sustained, weassessed cardiac function at 6 weeks and 12 weeks by ejection fractionand stroke volume using cardiac MRI. Ejection fraction of GFP-injected mice decreased to reach a stable value of approximately28% 6 weeks post-myocardial infarction (Fig. 4b). In contrast, infec-tion of injured myocardium with GHMT blunted worsening of ejec-tion fraction 6 weeks post-myocardial infarction (,49%) with furthersignificant improvement at 12 weeks post-myocardial infarction(,57%) (Fig. 4b). This long-term effect on ejection fraction byGHMT was accompanied by significant increases in stroke volumeat 12 weeks compared to 6 weeks (Fig. 4b). Individual mice in eachgroup demonstrated similar functional changes in both cardiac para-meters, indicating the reliability of cardiac MRI to assess cardiacfunction (Supplementary Fig. 19). These data indicate that expressionof GHMT in non-cardiomyocytes in injured hearts can sustaincardiac function. Moreover, GHMT- and GHMMsT-infected heartsshowed a pronounced reduction in fibrosis and increased muscletissue, compared with GFP-infected hearts after myocardial infarc-tion (Fig. 5a, b and Supplementary Fig. 20).

iCLM

End

og

eno

us

card

iom

yo

cyte

d

40 m

v

50 ms

0 mv

40 m

v

50 ms

0 mv

0

1

2

3

4

5

Tom

ato

+ c

ard

iom

yo

cyte

s

in inju

red

are

a (%

)

Uninjur

ed

Con

trol

GHM

T

MI

iCLM 1

iCLM 2

iCLM 3

iCLM 4

ba R26RtdT cTnT Merge

Inju

red

+ G

FP

Inju

red

+ G

HM

T

A

B

C

c

Figure 3 | Lineage tracing of GHMT-induced iCLMs in vivo. a, Border zoneof heart sections isolated from tamoxifen-treated Tcf21iCre/Rosa26RtdT micethat were subjected to LAD ligation followed by injection of GFP or GHMTretroviruses. Three weeks post-myocardial infarction, hearts were fixed,sectioned and stained for cTnT (green) and visualized for tomato (red). Upperpanels show GFP-injected heart in which tomato is seen only in fibroblasts.Lower panels show GHMT-injected heart in which tomato is seen in bothfibroblasts and cardiomyocytes. Three types of iCLMs (positive for cTnT andtomato) were observed in injured hearts injected with GHMT-virus: cell A,small cells without sarcomeres; cell B, has sarcomeres and expresses equivalentcTnT to neighbouring tomato-negative cardiomyocytes; cell C, has sarcomeresand expresses less cTnT than neighbouring tomato-negative cardiomyocytes.Scale bars, 20mm. b, Tomato1 iCLMs isolated from tamoxifen-treatedTcf21iCre/Rosa26RtdT mice 1 month post-myocardial infarction followed byinjection of GHMT retroviruses. Scale bar, 20mm. c, Quantification of tomato1

cardiomyocytes isolated from uninjured hearts (n 5 2), injured hearts treatedwith empty vector retroviruses (n 5 3), or GHMT retroviruses (n 5 3). Data arepresented as mean 6 s.d. d, Recording of action potential of tomato1 iCLMsand endogenous cardiomyocytes by patch-clamping. Action potentials wererecorded in response to brief (1–2 ms) depolarizing current (1–2 nA) injectionsdelivered at 1 Hz by whole-cell current patch-clamping. (n 5 5 for each group).

RESEARCH ARTICLE

6 0 2 | N A T U R E | V O L 4 8 5 | 3 1 M A Y 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

DiscussionOur results demonstrate that GATA4, HAND2, MEF2C and TBX5(GHMT) can reprogram cardiac fibroblasts into functional cardiac-like myocytes in vitro and in vivo, confirming and extendingprior studies16. Moreover, exogenous GHMT expression in non-cardiomyocytes of the heart post-myocardial infarction also reducesfibrosis and improves cardiac function.

Improvement of cardiac function post-myocardial infarction bydifferent transcription factor combinations correlated with their abilityto convert fibroblasts into iCLMs in vitro (Supplementary Fig. 21),suggesting that cardiac repair results, at least in part, from reprogram-ming of non-cardiomyocytes towards a cardiomyocyte fate. We feel it isimportant to emphasize, however, that the pronounced functionalimprovement of GHMT-injected hearts seems greater than might bepredicted from the relative inefficiency of reprogramming of adultcardiac fibroblasts to iCLMs in vitro. We speculate that the native milieuof the intact heart, containing extracellular matrix, growth factors, per-sistent contractility, surrounding contractile cells and other cell types, ismore permissive than plastic tissue culture dishes for functional repro-gramming. In this regard, reprogramming of pancreatic beta-cells hasbeen observed in vivo but not in vitro32. It is also conceivable that othermechanisms, such as a blockade to the activation of cardiac fibroblasts,enhanced survival of cardiomyocytes, facilitated differentiation ofactivated cardiac progenitors into cardiomyocytes, as describedpreviously27,31, or improved angiogenesis contribute to the benefits

observed upon expression of GHMT in the heart post-myocardialinfarction. Irrespective of these uncertainties, the strategy presentedhere, while still requiring optimization, provides a potential means ofimproving cardiac function in vivo, bypassing many of the obstaclesassociated with cellular transplantation.

METHODS SUMMARYAll experimental procedures with animals were approved by the InstitutionalAnimal Care and Use Committee at University of Texas Southwestern MedicalCenter.

Adult TTFs and cardiac fibroblasts were isolated by an explanting method inwhich fibroblasts migrate from minced tissue and grow in fibroblast growthmedium. Fibroblasts were transduced with a mixture of polybrene (Sigma; 6mgml21) and freshly made retroviruses expressing transcription factors generated inPlatinum E cells (Cell Biolabs). Twenty-four hours after viral transduction, theviral medium was changed to a cardiac induction medium. Medium was changedevery 2 days. Expression of cardiac genes was analysed by flow cytometry, micro-array and immunocytochemistry.

Adult mice, 8–10 weeks old, underwent either a sham operation or ligation ofthe LAD. Concentrated retroviruses (,108–9 infectious units (i.f.u.)) wereinjected into the border zone using a gastight 1710 syringe (Hamilton). Cardiac

b

Eje

ctio

n f

ractio

n (%

)

6 wee

ks

Uno

perat

ed

12 w

eeks

0

10

20

30

40

50

80

70

60

GFP (n = 10) Unoperated (n = 2)GHMT (n = 10)

****

Str

oke v

olu

me (m

l)

6 wee

ks

Uno

perat

ed

12 w

eeks

0

10

20

30

40

50

60

*

*

MIMI

ns*

**

ns

* **

aF

ractio

nal sh

ort

en

ing

(%

)

24 h

3 wee

ks

Sham

2 wee

ks0

10

20

30

40

50

60

MI MI

Eje

ctio

n f

ractio

n (%

)

24 h

3 wee

ks

Sham

2 wee

ks0

10

20

30

40

50

90

60

70

80

GFP (n = 9) GHMT (n = 10) Sham (n = 5)

Figure 4 | Ameliorated cardiac function of injured hearts by GHMT.a, Cardiac function of mice subjected to LAD ligation followed byintramyocardial injection of GFP or GHMT retroviruses was evaluated atvarious time points by echocardiography. Data are presented as mean 6 s.d.*P , 0.05; **P , 0.005; ns, not statistically significant. b, Cardiac function at 6and 12 weeks post-myocardial infarction was assessed by MRI. Data arepresented as mean 6 std. *P , 0.05, **P , 0.005.

b

Fib

rotic a

rea (%

)

GFP

GHM

T0

10

20

30

40

50

70**60

a

Mild

7/10

Moderate

2/10

Severe

1/10

GHMT

Mild

2/9

Moderate

2/9

GFP

Severe

5/9

L1

L2

L3

L4

L5

LV LV LV LV LV LV

500 μmL1L2L3L4L5

]

LV

80

0Severe Moderate Mild

30

50

70

10

40

60

20

Anim

al d

istr

ibutio

n (%

)

GHMTGFP

X

Figure 5 | Attenuation of fibrosis in response to myocardial infarction byGHMT. a, Comparison of cardiac fibrosis and scar formation between GFP-and GHMT-infected myocardium 4 weeks after LAD ligation. Cardiac fibrosiswas evaluated at five levels (L1–L5) by trichrome staining 4 weeks post-myocardial infarction. The ligation site is marked as X. Severity of cardiacfibrosis was classified as mild, moderate or severe (fibrotic area ,20%, 20–40%or .40%, respectively). Sections of representative hearts of each category ofseverity are shown. Numbers indicate the various type of severity/total numberof hearts examined in each group. A graph shows animal distribution in eachcategory. Scale bar, 1 mm. b, Quantification of fibrotic area in heart sectionsdisplayed in (a). Fibrotic area (%) 5 (the sum of fibrotic area at levels 3 and4/the sum of myocardial area in the LV at levels 3 and 4) 3 100. Data arepresented as mean 6 s.d. *P , 0.05, **P , 0.005.

ARTICLE RESEARCH

3 1 M A Y 2 0 1 2 | V O L 4 8 5 | N A T U R E | 6 0 3

Macmillan Publishers Limited. All rights reserved©2012

function was assessed using echocardiography and MRI. Hearts were collectedfrom euthanized animals for histological studies.

Isolation of cardiomyocytes, electrophysiological measurements, histology andimmunohistochemistry were performed as described previously2,33–37.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 18 March 2011; accepted 12 April 2012.

Published online 13 May 2012.

1. Lopez, A. D., Mathers, C. D., Ezzati, M., Jamison, D. T. & Murray, C. J. Global andregional burden of disease and risk factors, 2001: systematic analysis ofpopulation health data. Lancet 367, 1747–1757 (2006).

2. Porrello, E. R. et al. Transient regenerative potential of the neonatal mouse heart.Science 331, 1078–1080 (2011).

3. Yacoub, M., Suzuki, K. & Rosenthal, N. The future of regenerative therapy inpatients with chronic heart failure. Nat. Clin. Pract. Cardiovasc. Med. 3 (Suppl. 1),S133–S135 (2006).

4. Menasche, P. Cell-based therapy for heart disease: a clinically orientedperspective. Mol. Ther. 17, 758–766 (2009).

5. Hoover-Plow, J. & Gong, Y. Challenges for heart disease stem cell therapy. Vasc.Health Risk Manag. 8, 99–113 (2012).

6. Perin, E. C. et al. Effect of transendocardial delivery of autologous bone marrowmononuclear cells on functional capacity, left ventricular function, and perfusionin chronic heart failure: the FOCUS-CCTRN trial. J. Am. Med. Assoc., (2012).

7. Takahashi, K. & Yamanaka, S. Induction of pluripotent stem cells from mouseembryonic and adult fibroblast cultures by defined factors. Cell 126, 663–676(2006).

8. Yu, J. et al. Induced pluripotent stem cell lines derived from human somatic cells.Science 318, 1917–1920 (2007).

9. Tapscott, S. J. et al. MyoD1: a nuclear phosphoprotein requiring a Myc homologyregion to convert fibroblasts to myoblasts. Science 242, 405–411 (1988).

10. Wang, Z., Wang, D. Z., Pipes, G. C. & Olson, E. N. Myocardin is a master regulator ofsmooth muscle gene expression. Proc. Natl Acad. Sci. USA 100, 7129–7134(2003).

11. Vierbuchen, T. et al. Direct conversion of fibroblasts to functional neurons bydefined factors. Nature 463, 1035–1041 (2010).

12. Caiazzo, M. et al. Direct generation of functional dopaminergic neurons frommouse and human fibroblasts. Nature 476, 224–227 (2011).

13. Jugdutt, B. I. Ventricular remodeling after infarction and the extracellular collagenmatrix: when is enough enough? Circulation 108, 1395–1403 (2003).

14. Olson, E. N. Gene regulatory networks in the evolution and development of theheart. Science 313, 1922–1927 (2006).

15. Bondue, A. & Blanpain, C. Mesp1: a key regulator of cardiovascular lineagecommitment. Circ. Res. 107, 1414–1427 (2010).

16. Ieda, M. et al. Direct reprogramming of fibroblasts into functional cardiomyocytesby defined factors. Cell 142, 375–386 (2010).

17. Zang, M. X., Li, Y., Wang, H., Wang, J. B. & Jia, H. T. Cooperative interaction betweenthe basic helix-loop-helix transcription factor dHAND and myocyte enhancerfactor 2C regulates myocardial gene expression. J. Biol. Chem. 279, 54258–54263(2004).

18. Dai, Y. S., Cserjesi, P., Markham, B. E. & Molkentin, J. D. The transcription factorsGATA4 and dHAND physically interact to synergistically activate cardiac geneexpression through a p300-dependent mechanism. J. Biol. Chem. 277,24390–24398 (2002).

19. Garg, V. et al. GATA4 mutations cause human congenital heart defects and revealan interaction with TBX5. Nature 424, 443–447 (2003).

20. Ghosh, T. K. et al. Physical interaction between TBX5 and MEF2C is required forearly heart development. Mol. Cell. Biol. 29, 2205–2218 (2009).

21. Warren, L. et al. Highly efficient reprogramming to pluripotency and directeddifferentiation of human cells with synthetic modified mRNA. Cell Stem Cell 7,618–630 (2010).

22. Porter, K. E. & Turner, N. A. Cardiac fibroblasts: at the heart of myocardialremodeling. Pharmacol. Ther. 123, 255–278 (2009).

23. Byun, J. et al. Myocardial injury-induced fibroblast proliferation facilitatesretroviral-mediated gene transfer to the rat heart in vivo. J. Gene Med. 2, 2–10(2000).

24. Bhowmick, N. A. et al. TGF-b signaling in fibroblasts modulates the oncogenicpotential of adjacent epithelia. Science 303, 848–851 (2004).

25. Zeisberg, E. M. et al. Endothelial-to-mesenchymal transition contributes to cardiacfibrosis. Nature Med. 13, 952–961 (2007).

26. Schneider, M. et al. S100A4 is upregulated in injured myocardium and promotesgrowth and survival of cardiac myocytes. Cardiovasc. Res. 75, 40–50 (2007).

27. Loffredo, F. S. et al. Bone marrow-derived cell therapy stimulates endogenouscardiomyocyte progenitors and promotes cardiac repair. Cell Stem Cell 8,389–398 (2011).

28. Beyer, E. C. Gap junctions. Int. Rev. Cytol. 137C, 1–37 (1993).29. Saffitz, J. E., Green, K. G., Kraft, W. J., Schechtman, K. B. & Yamada, K. A. Effects of

diminished expression of connexin43 on gap junction number and size inventricular myocardium. Am. J. Physiol. Heart Circ. Physiol. 278, H1662–H1670(2000).

30. Acharya, A., Baek, S. T., Banfi, S., Eskiocak, B. & Tallquist, M. D. Efficient inducibleCre-mediated recombination in Tcf21 cell lineages in the heart and kidney.Genesis 49, 870–877 (2011).

31. Hsieh, P. C. et al. Evidence from a genetic fate-mapping study that stem cellsrefresh adult mammalian cardiomyocytes after injury. Nature Med. 13, 970–974(2007).

32. Zhou, Q., Brown, J., Kanarek, A., Rajagopal, J. & Melton, D. A. In vivo reprogrammingof adult pancreatic exocrine cells to b-cells. Nature 455, 627–632 (2008).

33. Tandan, S. et al. Physical and functional interaction between calcineurin and thecardiac L-type Ca21 channel. Circ. Res. 105, 51–60 (2009).

34. Laugwitz, K. L. et al. Postnatal isl11 cardioblasts enter fully differentiatedcardiomyocyte lineages. Nature 433, 647–653 (2005).

35. Luo, X. et al. Aberrant localization of intracellular organelles, Ca21 signaling, andexocytosis in Mist1 null mice. J. Biol. Chem. 280, 12668–12675 (2005).

36. Grynkiewicz, G., Poenie, M. & Tsien, R. Y. A new generation of Ca21 indicators withgreatly improved fluorescence properties. J. Biol. Chem. 260, 3440–3450 (1985).

37. Hardy, M. E. et al. Validation of a voltage-sensitive dye (di-4-ANEPPS)-basedmethod for assessing drug-induced delayed repolarisation in beagle dog leftventricular midmyocardial myocytes. J. Pharmacol. Toxicol. Methods 60, 94–106(2009).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank J. Cabrera for graphics. We are grateful to members ofthe Olson lab for critical reading of the manuscript. We thank D. Sosic, W. Tang, J.O’Rourke, N. Liu, M. Xin, A. Johnson and J. McAnally for discussions; J. Shelton andJ. A. Richardson for histology. We are grateful to I. Bezprozvanny for the PTI Ca21

Imaging System,andD.Srivastava for lentiviral plasmids.We thank the microarray coreatUniversity of Texas Southwestern MedicalCenter for collecting geneexpressiondata.E.N.O. is supported by grants from NIH, the Donald W. Reynolds Center for ClinicalCardiovascular Research, the Robert A. Welch Foundation (grant I-0025), the LeducqFondation-Transatlantic Network of Excellence in Cardiovascular Research Program,the American Heart Association-Jon Holden DeHaan Foundation and the CancerPrevention & Research Institute of Texas (CPRIT).

Author Contributions K.S. and E.N.O. conceived the project. K.S., Y.-J.N. and E.N.O.designed the experiments. K.S., Y.-J.N., X.L., X.Q., W.T., G.N.H., C.L.S. and A.A. performedexperiments. J.A.H. contributed to surgical experiments; E.G.N. made the Fsp1-Cremouse line. K.S. and R.B.-D. wrote the animal protocol. M.D.T. and R.B.-D. contributedscientific discussion. K.S., Y.-J.N., X.L., M.D.T., R.B.-D. and E.N.O. analysed data andprepared the manuscript.

Author Information Microarray data were deposited in the Gene Expression Omnibusdatabase (accession number GSE37057). Reprints and permissions information isavailable at www.nature.com/reprints. The authors declare no competing financialinterests. Readers are welcome to comment on the online version of this article atwww.nature.com/nature. Correspondence and requests for materials should beaddressed to E.N.O. (eric.olson@utsouthwestern.edu).

RESEARCH ARTICLE

6 0 4 | N A T U R E | V O L 4 8 5 | 3 1 M A Y 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

METHODSIsolation of primary fibroblasts. Tail-tip fibroblasts (TTFs): adult mouse tailswere skinned and cut into small pieces which were plated on tissue culture dishesand cultured in DMEM supplemented with 15% FBS and antibiotics. Themedium was changed every 2 to 3 days. TTFs migrated out from the explants after2 or 3 days. One week later, TTFs were frozen or replated for viral transduction.

Cardiac fibroblasts: hearts from adult mice (older than 4 weeks of age) wereminced into small pieces and plated on tissue culture dishes. Three minutes later,culture medium (DMEM:199 (4:1), 15% FBS and antibiotics) was gently added tothe dishes. Cardiac fibroblasts started to migrate out of the minced heart tissueafter 2 days. The medium was replaced every 2 days. Ten days later, cardiac fibro-blasts were frozen or replated for viral transduction.Generation of retroviruses. Retroviral plasmid DNA was generated by subclon-ing EGFP, Myc-tagged NKX2-5, GATA4, TBX5, HAND2, MEF2C and Flag-tagged MESP1 complementary DNAs into the retroviral vector pBabe-X38.

Ten micrograms of retroviral plasmid DNA was transfected using FuGENE 6(Roche) into Platinum E cells (Cell Biolabs) which were plated on a 10-cm tissueculture dish at a density of 33 106 cells per dish, 24 h before transfection. Twelvehours after transfection, medium was changed to 12 ml of fresh medium (DMEMsupplemented with 10% FBS and antibiotics). After 36 h of transfection, viral mediumwas harvested and filtered through a 0.45mm cellulose filter. The viral supernatantwas mixed with polybrene (Sigma) to a final concentration of 6mg ml21.Viral transduction. TTFs and cardiac fibroblasts were plated on tissue culturedishes pre-coated with SureCoat (Cellutron) at a density of 0.8 3 104 cm22. After24 h, the fibroblast growth medium was replaced with freshly made viral mixturecontaining polybrene. Twenty four hours later, viral medium was replaced withinduction medium, composed of 10% conditioned medium obtained fromneonatal rat/mouse cardiomyocyte culture, DMEM/199 (4:1), 10% FBS, 5% horseserum, antibiotics, non-essential amino acids, essential amino acids, B-27,insulin-selenium-transferrin, vitamin mixture, and sodium pyruvate(Invitrogen). Conditioned medium was filtered through a 0.22-mm pore sizecellulose filter. Medium was changed every 2 days until cells were collected.qPCR, western blot analysis and immunocytochemistry. Total RNA wasextracted from cultured cells and cDNA was synthesized by reverse transcription.All qPCR probes were obtained from Applied Biosystems. Western blots wereperformed with anti-Myc (Santa Cruz, clone A-14 1:1,000) and anti-Flag (Sigma,1:2,000) antibody. For immunocytochemistry, cells were fixed in 4% para-formaldehyde and incubated with primary antibodies: anti-GFP (Torrey PinesBiolabs 1:400), anti-cTnT (Thermo Scientific 1:400), anti-Myc (Santa Cruze,clone A-14 1:200), and anti-a-actinin (Sigma, 1:400). After washing with PBS,Alexa fluorogenic secondary antibodies (Invitrogen) were used to detect thesignal.DNA microarray. Total RNA was isolated from uninfected cardiac fibroblasts,cardiac fibroblasts transduced with either empty vector or GHMT retrovirusesand adult heart. Microarray analysis was performed on the platform of IlluminaMouse-6 Beadchip by the DNA Microarray Core Facility at the University ofTexas Southwestern Medical Center. Data were analysed using GeneSpring GXsoftware (Agilent).Flow cytometry. For the initial assay to detect aMHC-GFP expression, adherentfibroblasts were washed with PBS and detached from culture dish by treatmentwith Accutase (Millipore) for 10 min at 37 uC. Cells were then washed with 2% FBSin PBS and filtered through a cell strainer. Cells were incubated with propidiumiodide (1:1,000 dilution in 1% FBS in PBS) for 15 min at room temperature. Deadcells were excluded by propidium iodide staining and live cells were analysed forGFP expression using FACS Caliber (BD Sciences) and FlowJo software.

For intracellular staining of cardiac-specific markers, cells were fixed with 4%paraformaldehyde for 15 min after being harvested, as described above. Fixed cellswere washed with PBS and permeabilized with saponin for 10 min at room tem-perature. After being washed with PBS, cells were incubated with 5% goat anddonkey serum in PBS at room temperature for 30 min, followed by incubation withprimary antibodies (rabbit polyclonal anti-GFP antibody (Invitrogen) at a 1:100dilution and mouse monoclonal anti-cTnT antibody (Thermo Scientific) at a 1:400dilution in 0.2% goat and donkey serum in PBS for 30 min at room temperature.After washing twice with PBS, cells were incubated with secondary antibodies for30 min at room temperature. Secondary antibodies were goat anti-rabbit AlexaFluor 488 (Invitrogen) at a 1:200 dilution and donkey anti-mouse Cy5 (JacksonLaboratory) at a 1:400 dilution in PBS containing 0.2% goat and donkey serum.Cells were washed with PBS three times, and then analysed for GFP and cTnTexpression using FACS Caliber (BD Sciences) and FlowJo software.Generation of a-MHC-GFP transgenic mice. A plasmid containing a5.5 kilobase genomic fragment upstream of the mouse a-MHC gene plus exons1–3 and intronic sequences39, inserted upstream of a neomycin-resistance cassettefollowed by an internal ribosomal entry sequence (IRES) and a GFP reporter was

linearized and micro-injected into the pronucleus of zygotes. The founders werecrossed to C57BL/6 mice to get stable lines.Induction of Cre by tamoxifen administration. Tamoxifen (Sigma) was dis-solved in sesame oil (90%) and ethanol (10%) at a concentration of 50 mg ml21.To induce Cre activity, tamoxifen (0.2 mg g21 body weight) was administered bygavage with a 22-gauge feeding needle into mice bearing Tcf21iCre/1/R26RtdT ora-MHC-MerCreMer40/Rosa26-LacZ for three to five or seven consecutive days,respectively. Mice were analysed or subjected to myocardial infarction surgery atday 8 post-oral gavage of the last dosage.Sorting of Tcf21-expressing cells and gene expression. Three-month-oldTcf21iCre/1/R26RtdT mice were induced with 0.2 mg g21 tamoxifen for 3 con-secutive days by gavage, and 1 week later hearts were isolated and processed (atriaand aorto-pulmonary trunk were removed) to generate single cell suspensions forFACS sorting, as described previously41. The suspension was filtered throughtissue strainers, centrifuged at 400g for 5 min and resuspended in 10% CM media(10% Hyclone FBS, 3:1 DMEM/M-199, 10 mM HEPES, 1.2% antibiotic/antimycotic)before sorting with a MoFlo flow cytometer (Cytomation) using Summit software.For transcript analysis, sorted cells were collected into lysis buffer for RNA extrac-tion (RNAqueous Micro kit from Ambion). A fraction of each sample was alsocollected into PBS for post-sort assessment of purity. Complementary DNA wassynthesized using Superscript III reverse transcriptase (Invitrogen) and randomhexamers (Roche). Gene expression profiles were generated using standard qPCRmethods with iTAQ SYBR Green master mix (Bio-Rad) on a CFX96 instrument(Bio-Rad). Samples were run in triplicate and normalized to cyclophilin expres-sion. Fold enrichment was determined with respect to the unsorted population.Myocardial infarction surgery and intramyocardial injection of retroviruses.Mice were anesthetized with isoflurane, intubated with a polyethylene tube (size60), and then ventilated with a volume-cycled rodent respirator with 2–3 ml percycle at a respiratory rate of 120 cycles per min. Thoracotomy was performed atthe third intercostal space and self-retaining microretractors were placed to sepa-rate the third and fourth rib to visualize the LAD. A 7.0 prolene suture (Ethicon,Johnson & Johnson) was then passed under the LAD at 1.5 mm distal to the leftatrial appendage, immediately after the bifurcation of the left main coronaryartery. The LAD was doubly ligated. The occlusion was confirmed by the changeof colour (becoming paler) of the anterior wall of the left ventricle. Sham-operatedmice underwent the same procedure without ligation. Immediately after ligationof the LAD, 50ml of concentrated retrovirus (,108–109 infectious units (i.f.u.))was injected into the border zone of the infarct at five different areas using agastight 1710 syringe (Hamilton). The chest wall was then closed with a 5.0 Dexonabsorbable suture (Tyco Healthcare, United States Surgical), and the skin wasclosed with Topical Tissue Adhesive (Abbott Laboratory). Mice were extubatedand allowed to recover from surgery under a heating lamp. The mouse surgeonwas blinded to the study. Mice with fractional shortening .30% at day 1 post-myocardial infarction were removed from the study.Transthoracic echocardiography. Cardiac function was evaluated by two-dimensional transthoracic echocardiography on conscious mice using aVisualSonics Vevo2100 imaging system. Fractional shortening and ejection frac-tion were used as indices of cardiac contractile function. M-mode tracings wereused to measure left ventricular internal diameter at end diastole (LVIDd) andend systole (LVIDs). Fractional shortening (FS) was calculated according to thefollowing formula: FS (%) 5 [(LVIDd – LVIDs)/LVIDd] 3 100. Ejection fraction(EF) is estimated from [(LVEDV 2 LVESV)/LVEDV] 3 100%. LVESV, left vent-ricular end systolic volume; LVEDV, end diastolic volume. All measurementswere performed by an experienced operator blinded to the study.Cardiac MRI. Six and twelve weeks after myocardial infarction, the cardiacfunction of mice was re-evaluated by cardiac MRI using a 7 T small animal MRscanner (Varian) with a 38-mm birdcage RF coil. Under anaesthesia by inhalationof 1.5–2% isoflurane mixed with medical-grade air via nose-cone, the animalswere placed prone on a mouse sled, (Dazai Research Instruments) equipped witha pneumatic respiratory sensor and electrocardiogram (ECG) electrodes forcardiac sensing, head first with the heart centred with respect to the centre ofthe RF coil. The mouse chests were shaved and a conducting gel was applied tooptimize ECG contact between electrodes and mouse. All MRI acquisitions weregated using both cardiac and respiratory triggering. The bore temperature waskept at 35 uC to assure adequate and constant heart rate.

Two-dimensional (2D) gradient echo images on three orthogonal planes(transverse, coronal and sagittal) were acquired to determine the long-axis ofthe heart in each mouse. Axial images perpendicular to the long axis of the heartwere chosen for cine-imaging. Cine images at 12 phases per cardiac cycle wereobtained with an echo time of 2.75 ms, repetition time 5 ECG R-R interval /12,flip angle of 45u, and number of excitation (NEX) 5 4. Each scan consisted ofseven to ten contiguous slices from apex to LV outflow with 1-mm thickness, amatrix size of 128 3 128, and a field of view of 30 3 30 mm.

ARTICLE RESEARCH

Macmillan Publishers Limited. All rights reserved©2012

Epicardial and endocardial borders were manually traced for calculation ofLVESV and LVEDV using NIH ImageJ software. Total LV volumes were calcu-lated as the sum of all slice volumes. Stroke volume was calculated by the equa-tion, LVEDV 2 LVESV. Ejection fraction (EF) was calculated by the equation,[(LVEDV 2 LVESV)/LVEDV] 3 100%. Investigators performing MRI acquisi-tion and analysis were blinded to the assignment of mice group.Isolation of adult mouse cardiomyocytes, measurement of cardiomyocytesarcomere shortening and Ca21 transients. Mouse cardiomyocytes were iso-lated using enzymatic digestion and mechanical dispersion methods previouslydescribed33. In brief, after retrograde perfusion with Ca21-free Krebs–Ringerbuffer (KR, 35 mM NaCl, 4.75 mM KCl, 1.19 mM KH2P04, 16 mM Na2HPO4,134 mM sucrose, 25 mM NaCO3, 10 mM glucose, 10 mM HEPES, pH 7.4, withNaOH) and digestion with collagenase solution (collagenase II, 8 mg ml21), theventricular myocytes were separated using a fine scalpel and scissors. After gentletrituration, cells were kept in KB solution (10 mM taurine, 70 mM glutamic acid,25 mM KCl, 10 mM KH2PO4, 22 mM glucose, 0.5 mM EGTA, pH 7.2 with KOH)and studied within 6 h at room temperature.

To examine myocyte contractile capacity, isolated cardiomyocytes were incu-bated with 33 mM C12FDG for 30 min. The green C12FDG1 cardiomyocytes wereidentified by a fluorescence microscope. Adult cardiomyocytes from wild-typemice incubated with C12FDG were used to determine autofluorescence of cardi-omyocytes. C12FDG1 cardiomyocytes and C12FDG2 cardiomyocytes were field-stimulated at 1 Hz while being superfused with extracellular buffer at room tem-perature. Images were acquired at 240 Hz through a 360 microscope objectiveusing a variable field rate CCD camera (IonOptix). Cell length was measured by avideo edge-detection system, using an IonOptix interface system. IntracellularCa21 transients were measured as previously described33–36. Analyses were carriedout with the software (IonWizard, IonOptix). Only rod-shaped, clearly striatedcardiomyocytes that were Ca21-tolerant were used in the experiments.Calcium transient measurements in spontaneous beating cardiomyocytes.Calcium imaging of beating iCLMs, cultured neonatal mouse ventricular cardi-omyocytes was performed using the PTI (Photon Technology International)Ca21 Imaging System with an automated fluorescence microscope and a CCDcamera, as described previously35. Calcium transients in individual spontaneous

beating cell were calculated by measurement of Ca21-induced fluorescence atboth 340 and 380 nm.Membrane potential recordings of iCLMs in vitro. Action potentials of beatingiCLMs with di-4-ANEPPS (Invitrogen) were assessed, as described previously37.Action potential recordings of iCLMs isolated from adult mouse hearts.Tomato1 cardiomyocytes were identified under a Nikon inverted fluorescencemicroscope with a red filter. Whole-cell current clamp experiments (Axopatch200B, Molecular Devices) were conducted to measure cardiomyocyte membraneaction potentials. Cells were plated in a chamber that was superfused at 1 to2 ml min21 with standard solution A containing 140 mM NaCl, 5 mM KCl,1 mM MgCl2, 1 mM CaCl2, 10 mM HEPES (pH 7.4 with NaOH), and 10 mMglucose (310 mOsm). Recording pipettes were prepared with resistances of2–3 mV when filled with internal solution containing 135 mM KCl, 10 mMEGTA, 10 mM HEPES, 5 mM glucose (pH 7.2 with KOH; 300 mOsm). Actionpotentials were recorded in response to brief (1–2 ms) depolarizing current(1–2 nA) injections delivered at 1 Hz. All data were filtered at 5 kHz and analysedby pCLAMP 9 software (Molecular Devices).Histology and immunohistochemistry. Haematoxylin and eosin and LacZstaining were performed as described37. Frozen sections from ischaemic heartswere stained with anti-GFP (Aves Labs 1:800) antibody according to manufac-turer’s instructions. Paraffin-embedded sections were stained with anti-cTnT(1:400) and anti-Cx43 (Cell Signaling, 1:50). Frozen sections from tamoxifen-treated Tcf21iCre/1/R26RtdT mice (,5 week old) were stained with anti-P4HB(ProteinTech, 1:200), anti-cTnT (Thermo Scientific, 1:400), isolectin B4(Vector Labs, 1:100), and anti-SM22a (Abcam, 1:200). Signals were detected withAlexa fluorogenic secondary antibodies (Invitrogen).

38. Kitamura, T. et al. Efficient screening of retroviral cDNA expression libraries. Proc.Natl Acad. Sci. USA 92, 9146–9150 (1995).

39. Subramaniam, A. et al. Tissue-specific regulation of the a-myosin heavy chaingene promoter in transgenic mice. J. Biol. Chem. 266, 24613–24620 (1991).

40. Sohal, D. S. et al. Temporally regulated and tissue-specific gene manipulations inthe adult and embryonic heart using a tamoxifen-inducible Creprotein. Circ. Res.89, 20–25 (2001).

41. Russell, J. L. et al. A dynamic notch injury response activates epicardium andcontributes to fibrosis repair. Circ. Res. 108, 51–59 (2011).

RESEARCH ARTICLE

Macmillan Publishers Limited. All rights reserved©2012