Transcription Factors MYOCD, SRF, Mesp1 and SMARCD3 Enhance the Cardio-Inducing Effect of GATA4, TBX5, and MEF2C during Direct Cellular Reprogramming

Transient overexpression of defined combinations of master regulator genes can effectively induce cellular reprogramming: the acquisition of an alternative predicted phenotype from a differentiated cell lineage. This can be of particular importance in cardiac regenerative medicine wherein the heart lacks the capacity to heal itself, but simultaneously contains a large pool of fibroblasts. In this study we determined the cardio-inducing capacity of ten transcription factors to actuate cellular reprogramming of mouse embryonic fibroblasts into cardiomyocyte-like cells. Overexpression of transcription factors MYOCD and SRF alone or in conjunction with Mesp1 and SMARCD3 enhanced the basal but necessary cardio-inducing effect of the previously reported GATA4, TBX5, and MEF2C. In particular, combinations of five or seven transcription factors enhanced the activation of cardiac reporter vectors, and induced an upregulation of cardiac-specific genes. Global gene expression analysis also demonstrated a significantly greater cardio-inducing effect when the transcription factors MYOCD and SRF were used. Detection of cross-striated cells was highly dependent on the cell culture conditions and was enhanced by the addition of valproic acid and JAK inhibitor. Although we detected Ca2+ transient oscillations in the reprogrammed cells, we did not detect significant changes in resting membrane potential or spontaneously contracting cells. This study further elucidates the cardio-inducing effect of the transcriptional networks involved in cardiac cellular reprogramming, contributing to the ongoing rational design of a robust protocol required for cardiac regenerative therapies.


Introduction
Cardiovascular disease and ultimately heart failure resulting from myocardial infarction are caused mainly due to the lack of a sufficient number of cardiomyocytes -the contractile engine of the myocardium. Specifically, the average human adult left ventricle is comprised of 4 billion cardiomyocytes, and the sudden loss of approximately 25% of these cells due to an infarction injury leads to their eventual replacement with non-contractile scar tissue ultimately causing heart failure [1]. Over the past two decades significant effort has been made both in the lab and the clinic to prevent or even reverse heart failure through cell replacement and regeneration of the infarcted myocardium [2]. To fulfill the potential of such a therapeutic approach, a formidable challenge is to have available a sufficient number of de novo cardiomyocytes that can improve cardiac functional output [3].
Pluripotent embryonic stem (ES) cells, which theoretically have an unlimited expansion potential, can readily differentiate into spontaneously contracting cardiomyocytes, resembling the nascent myocardium in both function and electrophysiological properties [4,5]. Moreover, the discovery that transient overexpression of four transcription factors (TFs) is sufficient to epigenetically reprogram terminally differentiated somatic cells [6,7] into induced pluripotent stem (iPS) cells, which closely resemble ES cells and hold amongst the rest the same cardiac differentiation capacity [8,9] further revolutionized the field. This powerful methodology has also formed the foundation for a new scientific direction: ''direct epigenetic cellular reprogramming'' or ''induced transdifferentiation'' which can be generally defined as the acquisition of a distinct alternative cellular phenotype from a particular cell lineage [9][10][11].
In particular, Ieda et al. reported that simple overexpression of three genes (Gata4, Tbx5, and Mef2c) in neonatal cardiac and dermal mouse fibroblasts can result in induction of cardiomyocytelike cells in vitro [10]. Unsurprisingly, the three TFs used to achieve cellular reprogramming are positioned at the core of the genetic regulatory networks that govern developmental cardiogenesis across many evolutionary layers [12]. Two recent studies have reported similar results using either cardiac TFs or microRNA molecules [13,14], although a third study brought into question the capacity of Gata4, Tbx5, and Mef2c to achieve complete cellular reprogramming into cardiomyocytes [15]. Importantly, it has been recently demonstrated that intramyocardial viral delivery and overexpression of these cardiac TFs in the infarcted myocardium can induce in situ reprogramming of cardiac fibroblasts into cardiomyocytes, and improve the functional output of the heart [16,17]. Although significant strides have been made in this nascent field of epigenetic cardiac reprogramming, many issues require further investigation including the effect of additional cardiac TFs, the effect of the induction culture conditions, and the phenotypic characteristics of the reprogrammed cells.
In this study we evaluated the capacity of ten TFs to successfully induce cardiac cellular reprogramming of mouse embryonic fibroblasts (MEFs): Mm Mesp1 (M 1 ), Hs SMARCD3 (S 3 ), Hs MYOCD (M D ), Hs SRF (S F ), Hs NKX2-5 (N 5 ), Hs HAND1 (H 1 ), Hs HAND2 (H 2 ), Hs GATA4 (G 4 ), Hs TBX5 (T 5 ), Hs MEF2C (M C ). The TFs were distributed into four modules and various module combinations were used to transduce MEFs utilizing an inducibleexpression lentivirus-based gene delivery system. We established a set of criteria of increasing stringency to determine the cardioinducing effect and cellular reprogramming capacity of particular module combinations: 1) TF-induced binding and activation of cardiac-specific promoter elements, 2) Expression of endogenous cardiac-specific genes including those encoding for cardiac cytoskeletal proteins, cardiac transcription factors, calcium handling proteins and cardiomyocyte ion channel proteins, 3) Phenotypic cellular metamorphosis associated with cytoskeletal remodeling or reorganization, and in particular detection of a cross-striated cytoskeleton, and 4) Cardiomyocyte-like functional properties including resting membrane potential, intracellular calcium cycling, and spontaneous contractile activity. We determined that the modest cardiac reprogramming effect of GATA4, TBX5, and MEF2C could be enhanced with the addition of either MYOCD and SRF alone or in conjunction with Mesp1 and SMARCD3. Although transcription factor overexpression produced a strong and significant cardio-inducing effect, as determined by global gene expression analysis, formation of sarcomeres required fine-tuning of the culture conditions. Finally, we readily detected calcium transient cycling in reprogrammed cells although no significant differences were recorded when measuring the resting membrane potential.

Cell Culture and Reprogramming Induction
MEFs (Millipore, PMEF-HL & PMEF-NL), mouse NIH3T3 (ATCC, CRL-1658), and human HEK293T (ATCC, CRL-11268) were maintained and passaged in standard growth medium containing FBS (10%, Atlanta Biologicals), DMEM.HG (Life Technologies, 11960044), Glutamax (1X, Life Technologies, 35050058), Non-Essential Amino Acids (1X, Life Technologies, 11140050), Sodium Pyruvate (1X, Life Technologies, 11360070), bMercaptoethanol, (1X, Life Technologies, 21985023), and Gentamicin (25 mg/ml, Lonza, 17-518Z). Cells were passaged on tissue culture polystyrene pre-coated with porcine skin gelatin (Sigma, G1890). Transgenic MEFs isolated from mouse embryos (13 d.p.c.), were expanded in standard growth media conditions. To generate the transgenic animals the Myh6 promoter element was used to drive expression of eGFP and Puromycin N-acetyltransferase (PAC). The transgenic mice were maintained in a mixed 129Svev/C57BL/6 background. For the isolation of primary mouse embryonic fibroblasts, a single pregnant female mouse was euthanized by CO2 asphyxiation. Animal procedures were conducted according to the guidelines for the care and use of laboratory animals set and approved by the Institutional Animal Care & Use Committee (IACUC) of Duke University & Duke University Medical Center.
Primary cultures of MEFs isolated from the transgenic mice were epigenetically reprogrammed into induced pluripotent stem cells using a previously described polycistronic inducible-expression lentiviral vector [18] (FU.tet.on.OSKM). Briefly, following cell transduction with the lentiviral vector, cells were cultured in a reprogramming medium (20%FBS, DMEM.HG, NEAA, Sodium Pyruvate, Glutamax, bMercaptoethanol, Gentamicin, valproic acid, and leukemia inhibitory factor) in the presence of doxycycline. Doxycycline was removed 7 days post induction and individual colonies of induced pluripotent stem cells were mechanically picked and expanded. The cells were subsequently differentiated into cardiomyocytes using the previously described [19] hanging droplet technique for embryoid body formation.

DNA Vectors, Viral Particle Production, and Cell Transductions
DNA vectors designed for the production of lentiviral particles that allow the inducible expression of cardiac TFs in the targeted transduced cell type were assembled using a previously described lentiviral vector [20] 59 & 39 EcoRI). For the purposes of this project we chose to purchase only fully sequenced and validated cDNA clones from the Open Biosystems Mammalian Gene Collection. The human homolog of Mesp1 was not available when we were initiating this study and thus we purchased the mouse homolog and performed all our experiments with mouse Mesp1. To ensure the appropriateness of utilizing the human gene homologs we compared their protein sequence to that of their mouse counterparts using a multiple sequence alignment tool (http://www.ebi.ac.uk/Tools/msa/clustalw2/). Through the ClustalW2 protein sequence analysis we determined that there is a 100% homology in the functional protein domains (i.e. DNA binding domains) of both the mouse and human protein homologs. We used a previously described lentiviral inducible expression vector in order to reprogram MEFs into induced pluripotent stem cells: FU.tet.on.OSKM [18] (Addgene plasmid 20321).
To produce viral particles we used a previously described second generation lentivirus production system utilizing the psPAX2 (packaging vector, Addgene plasmid 12660) and pMD2.G (envelope vector, Addgene plasmid 12559) vectors (Dr. Didier Trono). We followed a procedure similar to the one described by Maherali N et al [20]. Briefly HEK293T cells were maintained in a standard growth medium and expanded in T-75 tissue culture flasks pre-coated with a 0.1% solution of porcine gelatin (Sigma). Cells were allowed to reach 90% confluence at which point they were transfected in the presence of Opti-MEMH (Life Technologies) with a total of 24 mg of the three lentiviral vectors (12 mg expression vector, 7.7 mg of psPAX2 and 4.3 mg of pMD2.G) using Lipofectamine 2000 (Life Technologies). The supernatant containing the viral particles was collected at 48 and 96 hours following initial transfection with a final volume of 20 ml. The supernatant was subsequently concentrated to a final volume of approximately 300 ml using Amicon Ultra-15 centrifugal filter unites (Millipore) and stored at 4uC for immediate use or in small aliquots at 280uC for long term use.
To transduce primary MEFs, the cells were plated at a density of approximately 10,000 cells/cm 2 is 6-well plates. The next day 1 ml of transduction medium containing 10 ml of each viral concentrate and 8 mg/ml sequabrene (Sigma) was used to transduce the cells. The same method was used to transduce the cell line NIH3T3 although we plated the cells at a concentration of 2,000 cells/cm 2 is the 6-well plates as they have a much higher proliferation rate. To ensure homogeneous transduction cells were first transduced with FUW.M2rtTA and the various reporter vectors, and subsequently passaged into new 6-well plates in order to be transduced with the various combinations of transcriptional modules. Transgene overexpression was induced by the addition of Doxycycline in the culture medium. Qualitative RT-PCR analysis was performed to ensure transcriptional expression of the delivered transgenes ( Figure S1).
Imaging, Fluorescence-activated Cell Sorting, and Immunofluorescence Fluorescent cell imaging was performed on either a Nikon Eclipse TE2000-U using a Roper Scientific CoolSnap HQ camera and the NIS Elements software suite or a Zeiss 510 inverted confocal microscope. FACS sorting was performed on either a DiVa sorter (Becton Dickinson) or a FACStar (BD Biosciences). FACS analysis was performed on a FACSCanto flow cytometer (BD Biosciences). For all the FACS experiments using the nontransgenic mouse embryonic fibroblasts, we used two cell populations as the negative control: negative control group 1 was comprised of fibroblasts transduced with only the FUW.M2rtTA and no reporter vector, and negative control group 2 was comprised of fibroblasts transduced with both the FUW.M2rtTA and one of the four reporter vectors. The gates which we defined for each of the reporter vectors were established as to not include any cells within negative control group 1. Metaanalysis of acquired FACS data was performed using Cyflogic (www.cyflogic.com) and Microsoft Excel. Primary antibodies used for immunofluorescence included: anti-Tnnt2 (R&D Systems, MAB1874), anti-Actn2 (Sigma, A7811), anti-Myh6 (Abcam, ab15), anti-Myl2 (Abcam, ab48003), anti-Acta2 (Abcam, ab5694), anti-Nppa (Abcam, ab14348), anti-Gja1 (Santa Cruz, SC-9059). Secondary antibodies used were raised in either chicken or goat against mouse or rabbit antibodies. They were purchased from Life Technologies and conjugated to Alexa FluorH 568 (red), Alexa FluorH 488 (green), or Alexa FluorH 350 (blue). Cell nuclei were detected using DAPI (Life Technologies, D1306). The effect of valproic acid addition on the level of cardio-induction was determined by counting Tnnt2 or Actn2 positive cells in the presence or absence of valproic acid.

Relative Gene Expression Analysis
Primers were designed using NCBI primer-BLAST. In order to avoid polymerization of non-specific DNA amplimers, when applicable primers were required to span an exon-exon junction and the primer pair was required to be separated by at least one intron on the corresponding genomic DNA (Table S1).
To perform qualitative RT.PCR analysis and ensure low level of DNA amplification without signal saturation, following total RNA isolation (Qiagen, 74104) a one-step RT.PCR kit (Qiagen, 21212) was utilized with 23 cycles of amplification for Gapdh and 26-28 cycles (BioRad, MyCycler) for the rest of the genes analyzed. Quantitative RT.PCR analysis was performed on either an ABI 7300 or 7900HT real time thermocycler using the QuantiTect SYBR Green one-step RT.PCR kit (Qiagen, 204243). While running the polymerase reactions on the real time thermocyclers we included the option of getting a dissociation curve and also run the final reaction product on agarose gels in order to ensure that the only amplimers detected and measured were the expected ones. The SDS software (ABI, version 1.4 or 2.4) was used to analyze the raw data and then additional analysis was performed on Microsoft Excel. Relative quantification was performed using the DDCt method and statistical significance was determined using the T-Test.
TaqMan Low Density Array Cards (TLDA) were custom designed using TaqMan probes (Table S1). Reverse transcription was performed using the high capacity RNA-to-cDNA kit (ABI, 4387496), and real time PCR was performed using the TaqMan Gene Expression Master Mix (ABI, 4369016) on an ABI 7900HT thermocycler (Duke genomics facility). Data analysis was performed using ABI RQ Manager 1.2.1, ABI SDS (2.4), and Microsoft Excel. Relative quantification was performed using the DDCt method and statistical significance was determined using the T-Test.

Electrophysiological Characterization
Sharp microelectrodes were fabricated from standard wall borosilicate glass capillary tubes (Sutter BF 100-50-10, Sutter Instruments) using a P-97 Sutter micropipette puller to generate electrodes with tip resistances between 50 and 70 MV when backfilled with 3 M KCl. The electrodes were connected to the headstage of a Multiclamp 700B amplifier (Axon Instruments, Inc.) using a silver chloride wire embedded within the micropipette holder. A reference silver chloride wire was connected to the bath chamber through an agar bridge. Cell cultures were perfused with warm (35-37uC) Tyrode's solution consisting of (mM) 135 NaC1, 5.4 KCl, 1.8 CaCl 2 , 1.0 MgCl 2 , 0.33 NaH 2 PO 4 , 5 HEPES, and 5 glucose; pH was adjusted to 7.4 with NaOH.
Membrane potential measurements were made using the current clamp mode of the Multiclamp 700B amplifier after electrode potential offset and capacitance were neutralized. Resting membrane potential was recorded from GFP(+) and GFP(2) cells (TNNT2.copGFP or TNNT2.GCaMP3) after the establishment of stable intracellular impalement. Data were sampled at 20 kHz and low-pass filtered using a 4.5 kHz Butterworth filter. Recordings were digitized using a BNC-2090 adaptor and NIDAQ-MX computer interface (National Instruments Corporation, Austin, TX). Signals were recorded and analyzed using WinWCP software (John Dempster, University of Strathclyde).
Videos of cells exhibiting GCaMP3 flashing were captured using a Nikon Eclipse TE2000-U with a Roper Scientific CoolSnap HQ camera and the NIS Elements software suite. The video files were subsequently analyzed with FIJI [26] as a stack of images to determine the average signal intensity in a region of interest within each cell. The average intensity for a given region of interest was normalized to the lowest value acquired and then plotted against the acquisition time of each frame. Spectral analysis was performed on linearly resampled (1 Hz) time series using Welch's method [27]. The 256-point fast Fourier transform was repeatedly computed with 50% overlap between adjacent segments. Then the spectral power of each segment was computed and averaged. Hanning window filtering was applied to avoid spectral leakage.

Microarray Gene Expression Analysis
MEFs were first transduced with the M2rtTA lentivirus and subsequently divided into four groups which were transduced with TF group 1 (G 4 T 5 M C ), TF group 2 (G 4 T 5 M C M 1 S 3 ), TF group 3 (G 4 T 5 M C M D S F M 1 S 3 ), or just FUW.M2rtTA (negative control). Induction of transgenic TF expression was initiated two days following lentiviral transduction at which point the standard MEF medium was supplemented with an additional 10%FBS, VPA (0.5 mM), and Dox (10 mg/ml). Total RNA was isolated 7 days following induction of TF expression (Qiagen, RNeasy) and expression of the TF was confirmed using qualitative RTPCR. 200 ng of RNA was prepared for microarray hybridization (Life Technologies, MessageAmp Premier RNA amplification kit). For the purposes of this experiment we utilized twelve (126) Affymetrix Mouse Genome 430A 2.0 microarrays (each cell group was repeated in triplicate) and the entire process of RNA amplification, hybridization, and raw data preparation was performed by the Duke University microarray core facility. The raw data files and experimental description have been submitted to NCBI Gene Expression Omnibus (GSE44401). For the heart and MEF control samples we acquired the CEL raw data files from NCBI Gene Expression Omnibus. Heart control (430 2. The main part of data analysis was performed using the Partek Genomics Suite. First, the raw data (CEL files) were imported and normalized using the RMA algorithm and subsequently ANOVA statistical analysis was performed on the entire data set searching for significant differences between any one of the three TF groups and the negative control. Subsequently significantly upregulated or downregulated genes were identified based on the fact that pvalue,0.05 and Fold Change, or .1.5. Hierarchical clustering analysis and self-organizing map analysis were performed within the software suite. For the purpose of integrating the data originating through Partek into genetic pathways, we performed pathway analysis by utilizing the Thomson Reuters Metacore GeneGo pathway suite. The Excel add-in statistical analysis suite XLSTAT was used to perform the principal component analysis.
We determined that when using the Nkx2-5 reporter, TF overexpression did not induce an increase in GFP(+) cells, while detected a significant GFP(+) cell fraction (3.3260.26%) in the negative control, indicative of the reporter vector's leakiness. No significant change was detected when measuring the gene expression levels of endogenous Nkx2-5. Data acquired using the Myl2 reporter closely resembled the levels of endogenous gene expression, although we detected a significant discrepancy when using a particular TF combination (G 4 T 5 M C M 1 S 3 ). This may be indicative of direct binding and activation of Myl2 transcription, by M 1 or S 3 . Background Myl2 reporter activity was lower as compared to that of the Nkx2-5 reporter (0.2860.07%). With the Myh6.eGFP reporter we detected a non-significant increase in the GFP(+) cell fraction for two TF combinations. The same two combinations produced a significant increase in endogenous Myh6 expression. Finally, the output measured when using the TNNT2.copGFP reporter resembled the endogenous Tnnt2 expression level for the two combinations containing MYOCD and SRF although we detected a significant GFP(+) cell fraction (22.3863.04%) in the negative control. A significant upregulation in Tnnt2 expression was detected with all four TF module combinations, although for two of them (same as for Myh6) the cardio-inducing effect was stronger (G 4 T 5 M C M D S F , and In addition to reporter vector activity, we also looked for morphological changes in the transduced cell populations ( Figure 1E). By day 7 we readily detected brighter and elongated GFP(+) cells clearly distinguishable from the rest of the GFP(+) cells. This observation was particularly evident in MEFs transduced with either To examine whether cardiac specific proteins were expressed in the transduced MEFs we performed immunofluorescence using antibodies against proteins Actn2, Tnnt2, Myh6, Myl2, and Acta2 (Positive control: Figure S3). At day 14 we detected a few positive cells in the group transduced with all the TF module combinations ( Figure 1F). The expressed proteins did not organize in a crossstriated cytoskeletal pattern usually detected in functional cardiomyocytes.
We also tested whether valproic acid, previously shown to increase the efficiency of iPS cell derivation [36], could affect the TF cardio-inducing effect. Simultaneous TF overexpression and valproic acid treatment significantly improved the overall cardioinducing effect by approximately two-fold as determined by the fraction of cells expressing either Actn2 or Tnnt2 (2.0760.51 fold, p-value: 0.004) ( Figure S4).
In addition to the morphological changes detected in a small cell fraction (elongation, and increased levels of GFP expression), we also observed a decrease in cell proliferation as compared to the negative control. Moreover, the TF-expressing cells and especially the ones transduced with the M D S F TF module, did not dissociate adequately with trypsin/EDTA and remained aggregated. Even when using a customized enzymatic solution, the dissociated cells did not attach once replated, and eventually underwent apoptosis. This was especially evident within the brightest GFP(+) cell fraction. We hypothesized that dissociation and expansion difficulties, coupled with the leaky activity of the reporter vectors were applying a negative selection on the cells undergoing reprogramming. To examine this we performed relative gene expression analysis on sorted and expanded GFP(2) and GFP(+) cells (Myh6.eGFP, Figure S5). We did not detect a significant upregulation of cardiac genes in the GFP(+) cell population further validating our hypothesis.
Expression of Mesp1, SMARCD3, MYOCD, and SRF Significantly Enhance the Cardio-inducing Effect of GATA4, TBX5, and MEF2C To further elucidate the cardio-inducing effect of the three TF modules while avoiding interference from the reporter vectors, we used MEFs from transgenic mice expressing GFP and puromycin N-acetyl-transferase (PAC) under the control of the cardiac mouse myosin heavy chain promoter [37]. First, we confirmed the lack of GFP expression in cultures of isolated MEFs, and tested whether the cells retained the capacity to express GFP in a cardiac-specific manner when epigenetically reprogrammed. Following iPS cell derivation and differentiation [18] [19], we confirmed specific GFP expression in spontaneously contracting cardiomyocytes ( Figure S6, Movie S1).
GFP expression was detected as early as 2 days post induction of TF overexpression in MEFs transduced with either G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 . By day 7 GFP expression was determined to be 1.6060.12% and 2.4060.11% respectively for the two TF groups (Figure 2A, B). This level of GFP expression was highly significant as compared to the negative control (0.0360.05%). Only rare events of GFP(+) cells were detected in MEFs transduced with only G 4 T 5 M C (0.0560.06%) and a slightly higher fraction in MEFs transduced with G 4 T 5 M C M 1 S 3 (0.2060.07%), which was not significant. It is noteworthy that we detected two main cell morphologies: small and round ( Figure 2C, lower) or large and elongated cells ( Figure 2C upper) which could be attributed to an effect caused by TF overexpression, or an influence from the surrounding cells.
We next examined whether GFP expression correlated with changes at the gene expression level of cardiac related genes ( Figure 2D, relative mRNA quantification). We custom-designed TaqMan Low Density Array plates allowing us to screen the transcriptional output of a large number of genes. Some of the overexpressed TF transgenes are usually active both during early cardiac development as well as in mature cardiomyocytes. The array plates were designed with the aim to detect both markers expressed in immature and mature cardiomyocytes. It is important to note that this analysis was performed on RNA isolated from the total cell population of each of the four TF groups and not from cells sorted based on GFP expression. This was done mainly because two of the TF groups (G 4 T 5 M C or G 4 T 5 M C M 1 S 3 ) did not induce any significant GFP expression in the transduced MEFs.
In general, using temporal gene expression analysis (7 or 14 days of TF overexpression), we recorded a significant effect in the transcriptional output (upregulation and downregulation) for some genes whereas for others no significant change was detected. This effect also varied amongst the different TF groups overexpressed. More specifically we detected a significant upregulation in the expression level of cardiomyocyte specific cytoskeletal proteins Actc1, Myh6, Myh7, Myl2, Tnni3, and Tnnt2. This effect was particularly evident in MEFs transduced with either G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 . The recorded differences also remained constant over time for most of the genes examined. When measuring the gene expression level of endogenous cardiac transcription factors we did not detect any large changes as compared to the control sample. It is noteworthy to mention that endogenous Mef2c expression was significantly increased, whereas Gata4 expression was slightly decreased, and Tbx5 expression was largely decreased for all TF groups used-. We also detected a significant decrease for Hand1 (Day 14) and Hand2 (Days 7 & 14), and a small, but significant increase for Nkx2-5 and Srf expression for two of the TF groups. Myocd was upregulated only in cells transduced with G 4 T 5 M C M 1 S 3 . Hop expression was significantly with G 4 T 5 M C , G 4 T 5 M C M 1 S 3 , G 4 T 5 M C M D S F , or G 4 T 5 M C M D S F M 1 S 3 . Negative control MEFs were only transduced with reporter vector and the M2rtTAexpressing lentivirus. The gates were defined using MEFs that were only transduced with the M2rtTA expressing lentivirus and no reporter vector. Following 7 days of induction of TF module expression, the fraction of cells expressing either GFP or mCherry was determined by FACS analysis. The relative gene expression level of the gene used in each of the reporter vectors was determined using quantitative RT.PCR analysis (endogenous locus). Results for both FACS and RTPCR are based on biological triplicates. Error bars represent calculated standard deviation (One *for p-value decreased in cells transduced with either G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 . The same analysis was also performed using wild-type MEFs transduced with three of the TF module groups and the results closely correlate with the ones presented in Figure 2 ( Figure S7).
We also examined the expression levels of genes that play a role in the electrophysiological function of cardiac myocytes. We did not detect any significant variations in the relative gene expression level for Atp2a1, Atp2a2, Cacna1a, Kcnj2, Kcnj3, Kcnk1, and Pln. Genes with significant expression upregulation included Casq2, Hcn4, and Nppa particularly in cells transduced with either The TF module G 4 T 5 M C had the strongest effect in inducing upregulation of Ryr2. We also detected a low level of Scn5a upregulation amongst all the TF groups, with G 4 T 5 M C M 1 S 3 being the most potent.
The TFs used are also highly expressed during cardiac development in cardiac progenitors [19,38], which have the capacity to differentiate into cardiomyocytes, smooth muscle and endothelial cells. When examining the expression levels of endothelial or smooth muscle specific genes, we detected small or no changes in the expression of Acta2, Kdr, Myl6, and Tagln. We detected a significant increase in expression of smooth muscle Myh11 in cells transduced with G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 . Finally, we detected a significant decrease in the expression level of fibroblast-specific genes Col1a1 and Col1a2 [17].
We also set out to determine the relative gene expression levels for the same genes described above when comparing positive control hearts to MEFs. Relative fold change for these genes was calculated using published Affymetrix microarray gene expression data (Table S2). The largest and most significant fold change was determined for genes Myh6, Myl2, Tnni3, Pln, Nppa, Ryr2, Tnnt2, Actc1, Casq2, Atp1a2, and Myl7. Significantly downregulated genes included Acta2, Tagln, Col1a2, and Col1a1. No significant change in gene expression was determined for Kcnj2, Kcnk1, Hand1, Hand2, Srf, Myl6, and Hcn4. Although the fold change determined when comparing cardiac tissue versus MEFs is significantly higher than the one determined when comparing the transduced MEFs versus the negative control it is noteworthy to mention that the RNA used in our experiments was isolated from the total population of transduced cells and not an enriched cell population.
We used antibodies against Actn2, Tnnt2, or Acta2 to determine whether the MEFs isolated from transgenic animals were expressing and organizing these proteins in cross-striated sarcomeres ( Figure 3A-C). We readily detected cells staining positive for all three of these proteins, however the staining pattern was not cross-striated indicating again an immature expression of these proteins. Actn2 and Tnnt2 expression did not always coincide with GFP expression, indicating, that some cells expressed these proteins without activating the Myh6.eGFP reporter locus.
We tested whether we could select the GFP(+) MEFs based on the fact that the same cardiac promoter element also controlled expression of PAC ( Figure 3D). We added a low concentration of puromycin in the culture medium and monitored the cells temporally. Almost immediately (Day 1-2) it became obvious that puromycin was toxic for the majority of GFP(2) cells, while the GFP(+) cells remained unaffected. We did not observe any proliferation in the GFP(+) cell population. On the contrary as more of the surrounding GFP(2) cells died, it led to the detachment of GFP(+) cells. This may be an indication that MEFs undergoing reprogramming require a supporting niche provided by the surrounding fibroblasts to survive. We also detected proliferation in GFP(+) MEFs transduced with either G 4 T 5 M C M 1 S 3 or only M 1 S 3 , suggesting that the combination of Mesp1 and SMARCD3 may be inducing transcriptional activation of the transgenic promoter element while at the same time inducing proliferation ( Figure 3E). Gene expression analysis performed on proliferative GFP(+) cells demonstrated a large and significant upregulation of endogenous Myl2.
Since the puromycin selection experiments suggested that the niche in which MEFs were being cultured may play an important role for their survival and possibly successful cellular reprogramming we tested whether establishing cocultures of pre-transduced MEFs with NRVMs would have an additional cardio-inducing effect. GFP(+) MEFs were readily detected within 2 days of induction of TF expression in cocultures when transduced with either G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 . At day 7 we detected three unique phenotypes for the GFP(+) cells: 1) the GFP was colocalized with either Actn2 or Tnnt2 and the two cytoskeletal proteins organized in a cross-striated pattern ( Figure 3G); 2) the GFP co-localized with either Actn2 or Tnnt2 but the two proteins were not organized in a cross-striated manner ( Figure 3H); and 3) the GFP did not co-localize with either of the two cardiac proteins ( Figure 3I). Connexin 43 (Gja1) that should be present along the borders of cardiomyocytes was not detected between the border of the transduced MEFs and NRVMs. We did not assay for fusion events between the GFP(+) cells and the NRVMs.

Global Transcriptome and Gene Pathway Analysis
To further ascertain the TFs' overall effect on global gene expression, we performed microarray gene expression analysis on MEFs transduced with: TF Group 1 (G 4 T 5 M C ), TF Group 2 (G 4 T 5 M C M 1 S 3 ), TF Group 3 (G 4 T 5 M C M D S F M 1 S 3 ), and negative control. The analysis was performed on the total population of MEFs. We first determined that the reprogrammed cells exhibited a variation in signal intensity for a number of probes ( Figure 4A). We selected probe sets exhibiting a significant level of transcript upregulation or downregulation (p-value: ,0.05, fold change,or .1.5) in each of the three groups as compared to the control ( Figure 4B, Table S3). By comparing the three sets of probes, we identified a core set of commonly upregulated (1065) or downregulated (980) genes ( Figure 4C-D, Table S4).
We implemented a pathway analysis tool to identify molecular pathways that were activated or inactivated in the cells undergoing epigenetic reprogramming. Previously described gene pathway networks received a calculated p-value based on the ratio of present-to-absent gene targets ( Figure 4E-G). The cardio-inducing Results are based on biological triplicates. Error bars represent calculated standard deviation (One *for p-value ,0.05, Two *for p-value ,0.01). C. GFP(+) MEFs were readily detected within 7 days of TF overexpression. We detected GFP(+) cells that were smaller in size (upper panels) or larger and elongated (lower panels). D. RNA was isolated at either 7 (black) or 14 days (gray) post induction of TF overexpression. Using quantitative RT.PCR and customdesigned TaqMan Low Density Array plates we measured the relative gene expression levels normalizing to negative control MEFs (FUW.M2rtTA only). Error bars represent calculated standard deviation (One *for p-value ,0.05, Two *for p-value ,0.01). doi:10.1371/journal.pone.0063577.g002 effect of TF group 3 was determined to be significantly higher as compared to that of TF group 1 or TF group 2 as it received significantly lower p-values for several cardiac-specific pathway networks. On the other hand, we detected a high level of alignment in all downregulated molecular pathways amongst all three TF groups. Notably downregulated pathways with the highest significance were related to some aspect of cell cycle control, indicating that cell division in general was significantly downregulated in the reprogrammed cells irrespective of the TF group applied and alluding to a common effect of GATA4, TBX5, and MEF2C.
Hierarchical clustering analysis performed on all probes exhibiting significant upregulation or downregulation in at least one of the three cell groups revealed particular probe set subgroups that differed in their level of gene expression ( Figure 4H). We hypothesized that the differences recorded across the three cell groups were the result of a varying effect on activation or repression of downstream targets. To further differentiate the cardio-inducing effect between the three TF groups, we performed self-organizing map clustering analysis and ultimately organized the probe sets into unique clusters for upregulated or downregulated genes ( Figure 4I-J, Table S5). We then performed pathway analysis on individual groups of clustered genes ( Figure S8). Upregulated clusters #1-8 containing genes with a lower expression level in TF group 3 received the lowest pvalues (highest significance) for network pathways including notch signaling, amyloid protein-based cell adhesion and ECM remodeling but no cardiac specific process networks. Upregulated Following induction of TF expression for 7 days we assayed cardiac protein expression (Actn2, Tnnt2) and gap junction formation (Gja1). In cocultures where MEFs were transduced with either G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 we detected three distinct staining patterns. Firstly, we detected GFP(+) cells with Actn2 or Tnnt2 co-localization. Importantly the cardiac proteins in these cells formed cross-striations (G). Secondly, we detected GFP(+) cells with Actn2 or Tnnt2 co-localization, where however the two cardiac proteins remained unorganized and no cross-striations where detected (H). Thirdly, we detected GFP(+) cells which were negative for either Actn2 or Tnnt2 expression (I). Gja1 staining did not indicate noticeable gap junction formation between the GFP-expressing cells and NRVMs. doi:10.1371/journal.pone.0063577.g003 clusters #15-22 containing genes with a higher expression level detected in TF group 3 received the lowest p-value for cardiac and muscle specific process networks including muscle contraction, skeletal muscle development, and cardiac development indicative of the significantly higher cardio-inducing potential of TF group 3 as compared to TF groups 1 and 2. Moreover, downregulated clusters 1-3, 6-7, 11, 12, and 15 containing genes with a lower expression level detected in cells transduced with TF group 3 received the lowest p-value for network pathways associated with cell cycle regulation. Downregulated clusters 13, 14, 18, 22, 23 containing genes with a higher expression level detected in cells transduced with TF group 3 received low p-values for network pathways including muscle and cardiac specific ones, further indicating the increased cardio-inducing potential of TF group 3.
Finally we evaluated the capacity of the three TF groups to induce cardiac cellular reprogramming in MEFs as compared to the effect described in a recent study reporting the successful in vitro reprogramming of cardiac fibroblasts into induced cardiomyocytes (iCM) [10]. We performed a network pathway analysis on genes selectively upregulated or downregulated in all three TF groups, heart control, and iCM ( Figure S9A-C). Although TF group 1 is comprised of the same set of TF as the ones used in this study (main difference being that we used the human gene orthologs), we observed the closest correlation for upregulated genes between our TF group 3 and their TF group with both being closest to the heart control as compared to TF group 1 and 2. Moreover, for the downregulated genes the iCM cells were the furthest away from the heart control indicative of the lack of correlation between the two.

Culture Conditions Enhance the Degree of Cytoskeletal Organization of Cardiac-specific Proteins
In addition to the evidence that short-term overexpression of certain TF module combinations led to transcriptional activation of cardiac-specific genes and expression of cardiac-specific proteins (although not properly organized), we examined whether the culture conditions would enhance this cardio-inducing effect. We screened three different media formulations including: a basic fibroblast growth medium (FGM), a cardiomyocyte growth medium (CGM), and a low serum growth medium (LSGM). We also examined whether extracellular matrix proteins may have a cardio-inducing effect by utilizing either gelatin or poly-l-lysine coated substrates [39]. Finally we examined the cardio-inducing effect of a JAK inhibitor which has been recently shown to enhance the cellular reprogramming capacity of mouse fibroblasts into cardiomyocytes [13].
We did not detect cross-striated Actn2(+)/Tnnt2(+) cells following culture of any of the four induced cell populations in either of the two culture medium formulations containing high serum concentrations (FGM or CGM). We detected rare events where cells expressed the two proteins without however organizing them in a cross-striated manner. When culturing the cells in LSGM, however, we readily detected Actn2(+)/Tnnt2(+) crossstriated cells ( Figure 5B, D, F, H). No Actn2(+)/Tnnt2(+) single or double-positive cells were detected in the negative control cell group ( Figure 5A). We determined that Tnnt2 expression always coincided with Actn2 expression whereas Actn2 was detected in Tnnt2(2) cells. Tnnt2 expression was detected in all cross-striated cells indicating that the particular marker is more specific for the detection of cardiomyocytes. Cross-striations were detected in all four cell groups, although we detected a significantly higher percentage of these cells when using either G 4 T 5 M C M D S F (0.7560.34%) or G 4 T 5 M C M D S F M 1 S 3 (0.5860.13%) ( Figure 5O). Additionally unlike the high level of cell proliferation observed when using the high serum media, very low levels of cell proliferation were detected when using LSGM and so both the density of the total number of cells as well as the density of the Tnnt2 positive cells was low even following 10 days of culture. We did not detect any significant cardio-inducing effect when culturing the cells on the two different extracellular matrix proteins. We did detect a small but significant cardio-inducing effect when using the JAK inhibitor leading to an average fold change of 1.7360.32 (p-value: 0.006) in the amount of Tnnt2 positive cells detected. We also detected double positive Actn2(+)/ Tnnt2(+) cells that did not organize the cytoskeletal proteins into cross-striated sarcomeres indicating that cellular reprogramming may have multiple stages of which not all are completed in target cells (Figure 5C, E, G, I). Over time we detected fewer crossstriated cells, and by day 30, the majority of double positive cells were not cross-striated indicating that the reprogrammed cells may require additional signaling for their long term survival and maintenance. Finally, we detected increased levels of Nppa protein expression in all four cell groups and in particular in the cells transduced with either G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 which closely correlates with the relative gene expression data reported earlier ( Figure 5J-N).

Electrophysiological Characterization of Reprogrammed Cells
We set out to examine the electrophysiological characteristics of MEFs transduced with the various combinations of TF modules. We first performed sharp electrode intracellular recording of resting membrane potential (RMP) in MEFs containing the TNNT2-copGFP reporter vector ( Figure 6A-B). NRVMs were used as positive control cells and their RMP was determined to be 275.7863.14 mV (n = 8). Although the TNNT2-copGFP reporter vector was previously shown to have a GFP expression pattern that closely resembled the transcriptional activation of the endogenous Tnnt2 gene locus ( Figure 1D), we chose to perform the recordings on both GFP(+) and GFP(2) cells to account for correspond to 1065 genes whereas the 1350 chip probes common for the significantly downregulated genes correspond to 980 genes (TF Group 1 (Green), TF Group 2 (Red), TF Group 3 (Blue)). E-G. Gene process networks that are either activated (E, upregulated genes) or deactivated (F, downregulated genes) in MEFs transduced with the three TF module combinations as compared to negative control, were determined using the Thomson Reuters GeneGo MetaCore TM data meta-analysis tool (TF Group 1 (Green), TF Group 2 (Red), TF Group 3 (Blue)). Based on the list of significantly upregulated or downregulated genes each process network received a p-value indicating the statistical probability that the network is affected in the population of reprogrammed cells. The range of calculated p-values for each process network is graphically represented with a green to red color range (G). Activated networks: Lowest p-value: 3.16610 29 (Green) and highest p-value: 6.53610 21   any leakiness of GFP expression. We did not detect any significant differences in RMP amongst the GFP(+) or GFP(2) cells for all experimental groups or when comparing GFP(+) to GFP(2) cells ( Figure 6B). We also recorded the RMP of GFP(+) transgenic mouse-derived MEFs (Myh6.eGFP) transduced with G 4 T 5 M C M D S F M 1 S 3 but their RMP (231.15610.46 mV, n = 12) was not significantly different from that recorded from negative control cells.
To investigate whether the recorded RMP in GFP(+) cells was influenced by potential intercellular coupling to surrounding cells via electrotonic loading [40], intracellular recordings were repeated on GFP(+) cells in the presence of 200 mM carbenoxolone, a gap junctional uncoupler [41]. Again no significant difference in RMP was observed in GFP(+) cells before (228.763.7 mV, n = 9) or after (227.364.2 mV, n = 6) the addition of carbenoxolone ( Figure 6C).
To determine whether the reprogrammed cells cycle cytoplasmic Ca 2+ , we transduced MEFs with a lentivirus that constitutively expresses the genetically encoded calcium indicator GCaMP3 [24,42]. Intracellular Ca 2+ concentration fluctuations can be detected as Ca 2+ binds to GCaMP3 and produces a transient increase in the intensity of its fluorescent signal.
Following, induction of TF module expression for 7 days we readily detected single cells or small groups of cells with a flashing GFP signal in MEFs transduced with either G 4 T 5 M C M D S F or G 4 T 5 M C M D S F M 1 S 3 ( Figure 6E, 6G, Movie S2, S4). We detected and recorded cells with fast and regular ( Figure 6I, L), or slower and irregular Ca 2+ transients ( Figure 6J, M). Furthermore, calculating and plotting the GCaMP3 signal spectral intensity allowed us to study the multiple frequencies incorporated in the intensity signal as well as identify the primary frequency of that signal. The range of the primary frequencies was determined to be between 0.31 Hz and 0.89 Hz for the two transcription factor combinations. We also detected rare events of Ca 2+ cycling in cells transduced with G 4 T 5 M C M 1 S 3 although the kinetics of Ca 2+ release within these cells was very slow, and significantly different from that observed for the other two transcriptional module combinations ( Figure 6K, Movie S3). We did not detect any changes in fluorescence intensity in the negative control cell population, indicating the absence of calcium cycling in those cells over an observation period of minutes ( Figure 6H).
Finally, we recorded the resting membrane potential of reprogrammed flashing cells. To control for potential leaky activity of the reporter vectors, we utilized a reporter vector in which GCaMP3 expression was controlled by the cardiac Troponin promoter. MEFs were either transduced with TNNT2.copGFP or TNNT2.GCaMP3 and RMP measurements were recorded from either GFP(+) or GCaMP3(+) flashing cells ( Figure 6D). No significant difference was detected in the membrane potential of the two cell groups when transduced with either of the transcriptional modules G 4 T 5 M C M 1 S 3 or G 4 T 5 M C M D S F M 1 S 3 .

Discussion
Here we describe a systematic study to determine the capacity of ten transcription factors to induce a cardiomyocyte-like phenotype in cultures of MEFs. We demonstrate that TFs M D S F alone or in conjunction with M 1 S 3 significantly enhance the basal but indispensable cardio-inducing effect of G 4 T 5 M C . More specifically, when we overexpressed the two groups of either five or seven TF we detected: 1) TF-induced binding and activation of cardiac-specific promoter elements, 2) Expression of endogenous cardiac-specific genes including those encoding for cardiac cytoskeletal proteins and cardiac transcription factors. We also detected a mixed effect for calcium handling genes and cardiomyocyte ion-channel genes as some were upregulated whereas others were not changed. 3) Phenotypic cellular metamorphosis associated with cytoskeletal remodeling or reorganization and in particular detection of cross-striated cytoskeletal proteins, and 4) Calcium transient oscillations, although we did not detect significant changes in resting membrane potential or presence of contractile activity.
Interestingly we did not detect a cardio-inducing effect when transducing MEFs with either of the two TF pairs alone or together in the absence of G 4 T 5 M C signifying the importance of the latter as a necessary component in the cardio-inducing process [10]. Although this effect was evident when using lentivirallydelivered reporter vectors (Myl2.mCherry, TNNT2.copGFP), we detected a significant fraction of GFP(+) or mCherry(+) cells in the negative control, an observation likely related to the manner of reporter vector delivery, and indicative of a potential drawback in utilizing such an approach in detecting cellular reprogramming events accurately. However, when utilizing MEFs isolated from a transgenic mouse we readily detected a significant increase in the fraction of cells expressing GFP when using either only M D S F (1.6060.12%) or in conjunction with M 1 S 3 (2.4060.11%) as compared to the negative control (0.0360.05%) or when just using G 4 T 5 M C (0.0560.06%). Three groups studying the overexpression of Gata4, Tbx5, and Mef2c (mouse orthologs) in mouse cardiac fibroblasts, mouse tail-tip fibroblasts, or mouse embryonic fibroblasts have reported approximately 20% [10], 2.2% [14], and 0% [15] activation of a fluorescent marker controlled by the Myh6 promoter element. The reasons for this large discrepancy may include differences in experimental procedures, cell and virus preparation, or even timing of TF overexpression [14,15]. In our study we utilized the human orthologs of the delivered genes (except for Mesp1), which may have affected the efficiency of cellular reprogramming.
We also detected a significant cardio-inducing enhancement when adding TF M D S F alone or with M 1 S 3 and both in conjunction with G 4 T 5 M C as determined by gene expression analysis. This effect was highly significant for Actc1, Myh6, Myl2, Tnnt2, Casq2, Hcn4, Srf, Acta2, Nppa, and Myh11. In the absence of M D S F , the previously detected cardio-inducing effect was eliminated when using only G 4 T 5 M C or in conjunction with M 1 S 3 . This may indicate that M D S F alone enhance the cardio-inducing effect of G 4 T 5 M C for the genes listed above whereas addition of M 1 S 3 alone does not significantly alter their expression levels. The strong cardio-inducing effect of M D S F in conjunction with G 4 T 5 M C may be due to the fact that both genes are active and necessary during embryonic cardiogenesis [43][44][45]. Interestingly addition of M D S F in conjunction with G 4 T 5 M C negated the upregulation recorded for the expression level of endogenous Myocd, which was only significantly upregulated when M 1 S 3 was used, whereas it was significantly downregulated in the other conditions. Additionally although overexpression of G 4 T 5 M C alone or with M 1 S 3 induced a significant downregulation in the relative expression levels of Nkx2-5 and Hop, M D S F addition rescued this effect which is noteworthy since it has previously been shown that Hop expression is regulated by Nkx2-5 and importantly it is an antagonist of Srf regulating the amount of cardiomyocytes in the developing heart [46,47]. Future experiments may involve testing the cardio-inducing effect of MYOCD alone in the absence of SRF and more importantly in the presence of overexpressed HOP. It is also noteworthy to mention that endogenous Tbx5, and Hand2 expression was significantly downregulated in all groups indicating that they are negatively regulated by the combination of G 4 T 5 M C . Based on our . ANOVA was used to determine whether significant differences existed in the measured membrane resting potentials amongst the reported cell groups or between the GFP(+) and GFP(2) cell populations within each cell group. (One *for p-value ,0.05, Two *for p-value ,0.01). NRVMs as positive control (RMP: 275.7863.14 mV, n = 8). Error bars represent calculated standard deviation. C. Sharp microelectrode recordings of RMP performed on GFP(+) experimental observations G 4 T 5 M C expression alone had a generally weak cardio-inducing effect, although we detected a significant upregulation of Tnnt2 expression as previously reported [15]. Also, the weak or absent activation of transcriptional expression of smooth muscle or endothelial markers suggests that the cardiac reprogramming of fibroblasts does not proceed through the formation of a multipotent cardiac progenitor cell [10,19]. Finally, our gene expression data suggested that both ventricular and atrial cardiomyocyte-like subtypes were present in the cells undergoing epigenetic reprogramming as evident by expression of Myl2, Myl7 and Nppa. In our future studies we plan to further characterize these cellular subtypes using electrophysiogical characterization.
The enhanced cardio-inducing effect of the TF pair M D S F in conjunction with G 4 T 5 M C was particularly evident when we performed global gene expression analysis. Cardiac development and muscle-function gene process networks identified based on global upregulated gene expression received significantly lower pvalues for cells which were also transduced with M D S F . We detected activation of gene process networks that are associated with cell adhesion, which may explain why the MEFs could not be easily enzymatically dissociated following induction of TF expression. On the other hand, when analyzing downregulated genes the global effect on gene expression was homogeneous and highly aligned in all three TF groups used, indicative of a role of the common denominator transcription factor module: G 4 T 5 M C . More importantly the gene process networks that were determined to be significantly downregulated are associated with aspects of the cell cycle. This finding is aligned with our observation that we detected a lack of proliferation, which is to be expected if the MEFs were being reprogrammed into cardiomyocytes [48]. It has recently been demonstrated that the number of cell divisions during the derivation of induced pluripotent stem cells is a fundamental variable driving epigenetic reprogramming to pluripotency with every cell being capable of eventually being reprogrammed into an iPS cell [49]. This inherently may create a significant limitation for the potential of increasing the efficiency of cardiac cell reprogramming, as the TF used in this study inhibit cell proliferation and may in turn limit their access to the binding sites in the genome. This may also explain the fact that we detected so many immaturely reprogrammed cells lacking cytoskeletal organization or proper electrophysiological properties. Finally, using global gene expression analysis we determined that MEFs transduced with G 4 T 5 M C M D S F M 1 S 3 were more similar with the induced cardiomyocytes derived using Gata4, Tbx5, and Mef2c [10], an observation aligned with the FACS data indicating potential differences in the cell source used, the method of gene delivery, or the human versus mouse gene orthologs. In the future we plan to include within this analysis gene expression profiles of unique cardiac resident cell subtypes including ventricular or atrial cardiomyocytes, pacemaker cells, cardiac fibroblasts, smooth muscle, and endothelial cells. Such an analysis would better elucidate the gene expression signature of the reprogrammed cells.
Addition of small molecules significantly enhanced the fraction of cells expressing cardiac proteins. Valproic acid, a histone deacetylase inhibitor, previously shown to significantly improve the efficiency of pluripotency induction [36,44], provided a twofold increase of the cardio-inducing effect in our experiments. Furthermore, use of the JAK inhibitor which was previously shown to have pro-cardiogenic activity [13,50], also significantly enhanced the cardio-inducing effect of the delivered TF. Although addition of these two small molecules significantly increased the fraction of cells expressing cardiac specific proteins, organization of these proteins in a cross-striated manner was only depended on cells being cultured in the low-serum growth medium. This suggests that the means of action of these two molecules is to increase TF accessibility to their DNA binding sites, or to prime the cells for reprogramming, thus increasing the probability that the target cells initiate cardiac protein expression. On the other hand, culture of the cells undergoing reprogramming in low serum growth medium seems to be the determinant of cytoskeletal maturation. In general culture of MEFs under low serum conditions significantly decreases their proliferative capacity and since the TF used in this study cause a significant downregulation of cell cycling it is possible that the low serum culture conditions enhance this effect. Further studies are necessary to define the effect of the culture conditions on initiation of reprogramming and maturation of the desired cells. Additionally, we would like to optimize the culture conditions with the ultimate goal of significantly improving the completeness and efficiency of reprogramming fibroblasts into cardiomyocytes [51]. Such an approach should include optimizing the time window during which the TF are being overexpressed. It should also include determining the ideal TF stoichiometry which has been previously shown to influence both the state and the properties of the derived cells [52].
We did not detect a significant hyperpolarization of the RMP in MEFs transduced with any of the TF combinations up to 7 days following induction of expression. Previously the RMP for wildtype mouse cardiomyocytes was reported to be approximately 256 to 263 mV [53,54] and a similar RMP (257 mV) was recently reported in a study describing the reprogramming of mouse fibroblasts into cardiomyocytes [10]. Moreover, hyperpolarization of RMP to a level of that expected to be recorded in cardiomyocytes was only detected following long-term cell culture which may indicate that the proteins responsible for RMP hyperpolarization are only expressed in later stages of reprogramming. However in our study long-term cell culture caused loss of cross-striated sarcomeres and selective proliferation of nontransduced fibroblasts. This is consistent with our observation that we did not detect significant upregulation in genes conferring electrophysiological function including Atp2a1, Atp2a2, Cacna1a, Kcnj2, Kcnj3, Kcnk1, Pln, and Scn5a. The recorded RMPs in both the GFP(+) and GFP(2) MEFs more closely resembled those observed in other unexcitable cells including mesenchymal stem cells (219 to 235 mV) [55,56], skeletal myoblasts (226 to 244 mV) [57,58], and cardiac fibroblasts (220 to 237 mV) [59,60]. On the other hand, we readily detected increased incidence of intracellular Ca 2+ oscillations using the genetically encoded calcium indicator GCaMP3 [24,42], particularly in MEFs transduced with M D S F alone or with M 1 S 3 in conjunction with G 4 T 5 M C . This observation is in agreement with recent reports describing the detection of spontaneous Ca 2+ oscillations of variable frequency in reprogrammed mouse fibroblasts [10,15]. Although we detected Ca 2+ oscillations in cells transduced with only G 4 T 5 M C M 1 S 3 , the intracellular calcium waves were significantly slower as compared to those transduced with G 4 T 5 M C M D S F M 1 S 3 . Overall, based on the depolarized resting potential that would preclude Ca 2+ flux through membranebound, voltage-dependent Ca 2+ channels, the observed intracellular Ca 2+ transients likely originated from cyclical oscillations of Ca 2+ levels in intracellular calcium stores.
A reemerging observation evident throughout our study is the strong enhancement of the cardio-inductive effect of G 4 T 5 M C achieved by the addition of M D S F . Myocardin is a strong transcriptional coactivator expressed in cardiomyocytes and smooth muscle cells during postnatal development, and along with Mkl1 and Mkl2 it associates with Srf which binds on CArG DNA motifs activating transcription [61]. Forced Myocd overexpression has been shown to activate expression of Acta2, Tagln, and Myh11 [43,62]. Importantly, Myocd is required for the maintenance of heart function by maintaining sarcomeric organization and intercalated disc structures, and promoting cardiomyocyte survival. Although Myocd expression has not been detected in cardiac fibroblasts thus far, its cofactor Mlk1 is indeed expressed and contributes to the induction of the myofibroblast phenotype following myocardial infarction injury [63]. Two recent reports demonstrated that following myocardial infarction delivery of Gata4, Mef2c, and Tbx5, or GATA4, HAND2, MEF2C, and TBX5 in the injured myocardium successfully reprogrammed cardiac fibroblasts into cardiomyocytes [16,17]. In these studies the authors report that the in vivo derived cells were more fully reprogrammed, more closely resembled host cardiomyocytes, and that TF overexpression induced a functional improvement that was greater to that predicted based on the in vitro reprogramming efficiency of the same TF combination. We hypothesize that, based on our observation when using the M D S F transcriptional module, that the reported in vivo effects may be the result of transduction and cellular reprogramming of activated myofibroblasts present in the infarcted region, which although may be lacking Myocd expression, have activated the Srf/Mlk1 transcriptional pathways. Additional experiments will need to be performed with fibroblasts and activated myofibroblasts to test whether the activation process itself would enhance the cardiac reprogramming efficiency.
In conclusion, here we describe a detailed study to determine the capacity of a core set of transcription factors to induce cellular reprogramming of primary fibroblasts into cardiomyocytes. We demonstrate that MYOCD and SRF alone or in conjunction with Mesp1 and SMARCD3 significantly enhance the cardio-inducing effect of GATA4, TBX5, and MEF2C. We also show that derivation of cardiomyocyte-like cells containing well-organized cross-striated sarcomeres is highly dependent on the culture conditions used during reprogramming. The reprogrammed cells develop the capacity to cycle intracellular Ca 2+ , although we do not detect significant membrane hyperpolarization or spontaneous contractions. It is evident that the complex genetic networks active during embryonic cardiac development can also induce cardiac cell reprogramming of non-cardiac cell types although this process is currently inefficient and poorly understood. Our work sheds light into the role of some of the main genetic regulators participating in this process. The effect of the combinatorial TF module overexpression on the endogenous level of Myh6 transcription did not correlate with the effect measured using the Myh6.eGFP reporter vector. In the case of Myl2 transcription, the only TF module that induced a large and significant upregulation effect was M 1 S 3 . That effect was abolished when M 1 S 3 was co-expressed with any of the other three TF modules. Finally, a much clearer and significant transcriptional effect was established when measuring the relative expression level of the endogenous Tnnt2 which suggested a strong effect of the G 4 T 5 M C TF module in its activation. A strong Tnnt2 expression as a result of G 4 T 5 M C was consistent in both the NIH3T3 and primary MEFs. The effect of G 4 T 5 M C on the level of Tnnt2 expression was more profound when using NIH3T3 (100-1000X), which may indicate an increased epigenetic accessibility of the overexpressed TF in the immortalized cell line as compared to primary MEFs. Error bars represent calculated standard deviation. (One *for p-value ,0.05, Two *for p-value ,0.01). Error bars represent calculated standard deviation. (TIF) Figure S3 Positive control immunofluorescence staining. Primary cultures of isolated neonatal rat ventricular myocytes (NRVMs) were stained using antibodies raised against the Actn2, Tnnt2, Myh6, Myl2, Gja1, and Acta2 cardiac proteins. Following rat ventricular tissue digestion, cardiac fibroblasts were removed following two cycles of pre-plating and subsequently NRVMs were plated on fibronectin-coated cell culture plates. Fixation and immunofluorescent analysis was performed 7 days following initial plating and culture. Cells were spontaneously contracting in synchrony prior to fixation.  Figure S6 Confirmation of GFP expression in a cardiacspecific manner in spontaneously contracting cardiomyocytes derived from differentiated induced pluripotent stem cells derived from primary transgenic MEFs (Myh6.eGFP.Myh6.PAC). A. We used a previously described [18] lentiviral vector (FU.tet.on.OSKM) to successfully reprogram primary MEFs isolated from the transgenic mice. Colonies of undifferentiated iPS cells were readily detectable within 10 days following induction of expression of Oct4, Sox2, Klf4, and Myc. Colonies were mechanically picked, enzymatically dissociated, and passaged on feeder layers of mitotically-inactivated primary MEFs. B. We used the previously described [19] hanging droplet technique to differentiate the derived iPSCs. Within 7 days of differentiation we readily detected spontaneously contracting population of GFP-expressing cardiomyocytes. This allowed us to confirm that the transgenic MEFs retain their capacity to express GFP in a cardiac specific manner even when undergoing a cycle of reprogramming into undifferentiated iPSC and then being differentiated again. (TIF) Figure S7 Gene expression analysis of primary MEFs transduced with three combinations of transcriptional modules. MEFs were transduced with FUW.M2rtTA and G 4 T 5 M C , or G 4 T 5 M C M 1 S 3 , or G 4 T 5 M C M D S F M 1 S 3 . We isolated RNA on 7 days post induction of transcription factor expression. Using quantitative RT.PCR and custom-designed TaqMan Low Density Array plates we measured the relative gene expression levels normalizing to negative control MEFs (FUW.M2rtTA only). (One *for p-value ,0.05, Two *for p-value ,0.01). Error bars represent calculated standard deviation. (TIF) Figure S8 Genes belonging to groups of gene clusters showing similar patterns were further analyzed using the Thomson Reuters GeneGo MetaCore TM data metaanalysis tool: clusters 1-8 (upregulated), clusters 15-22 (upregulated), clusters 1-3 & 6-7 (downregulated), clusters 11, 12, 15 (downregulated), clusters 13, 14, 18, 22, 23 (downregulated). Process networks were determined for each of the groups based on a p-value score with the most significant having the lowest p-value. (TIF) Figure S9 Microarray gene expression analysis performed on populations of transdifferentiated MEFs, Heart Control, and induced cardiomyocytes (iCMs). A-C. Gene process networks that are either activated (A, upregulated genes) or deactivated (B, downregulated genes) in MEFs transduced with the three transcriptional module combinations as compared to negative control, heart positive control as compared to MEFs negative control, and iCMs as compared to cardiac fibroblasts negative control, were determined using the Thomson Reuters GeneGo MetaCore TM data meta-analysis tool (TF Group 1 Green, TF Group 2 Red, TF Group 3 Blue, Heart Control Black, iCMs Doted line). Based on the list of significantly upregulated or downregulated genes each process network received a p-value indicating the statistical probability that the network is affected in the population of transdifferentiated cells, or heart control. The range of calculated p-values for each process network is graphically represented with a green to red color range (C). Activated networks: Lowest p-value: 4.12610 218 (Green) and highest p-value: 9.81610 21 (Red). Deactivated networks: Lowest p-value: 5.17610 241 and highest p-value: 1 (Red). D-E. Principal component analysis was performed on the p-values calculated for each of the process networks for either significantly upregulated (D), or significantly downregulated genes (E). Correlation circle plots are graphically represented for each of the three transcription factor groups, the iCMs, and the heart. (TIF) Table S1 Lists of primers and Taqman probe IDs used for qualitative and quantitative RT.PCR analysis as well as the ones used to design the custom-made TLDA plates.

(XLSX)
Table S2 Calculated fold change for genes that are significantly upregulated or downregulated when comparing the transcriptome of heart (positive control) to that of MEFs (negative control). The genes in this list are the same presented in Figure 2.