Author contributions: H.H. and N.K.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript; T.T.: conception and design, collection and assembly of data, data analysis and, interpretation, final approval of manuscript; N.K.-K., S.K., and P.K.: collection and assembly of data, final approval of manuscript; M.F.: financial support, data analysis and interpretation, final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS July 5, 2011.
Induced pluripotent stem cells (iPSCs) can be created by reprogramming differentiated cells through introduction of defined genes, most commonly Oct4, Sox2, Klf4, and c-Myc (OSKM). However, this process is slow and extremely inefficient. Here, we demonstrate radical acceleration of iPSC creation with a fusion gene between Oct4 and the powerful transactivation domain (TAD) of MyoD (M3O). Transduction of M3O as well as Sox2, Klf4, and c-Myc into fibroblasts effectively remodeled patterns of DNA methylation, chromatin accessibility, histone modifications, and protein binding at pluripotency genes, raising the efficiency of making mouse and human iPSCs more than 50-fold in comparison to OSKM. These results identified that one of the most critical barriers to iPSC creation is poor chromatin accessibility and protein recruitment to pluripotency genes. The MyoD TAD has a capability of overcoming this problem. Our approach of fusing TADs to unrelated transcription factors has far-reaching implications as a powerful tool for transcriptional reprogramming beyond application to iPSC technology. STEM CELLS 2011; 29:1349–1361
Creation of induced pluripotent stem cells (iPSCs) can be achieved through a surprisingly simple approach to nuclear reprogramming . In the original protocol, Takahashi and Yamanaka transduced retrovirus encoding Oct4, Sox2, Klf4, and c-Myc (OSKM) into mouse embryonic fibroblasts (MEFs) and obtained pluripotent cell colonies resembling embryonic stem cells (ESCs) in approximately 2 weeks . However, iPSC protocols are generally very inefficient, with only around 0.1% of MEFs transduced with OSKM forming iPSCs. Although the reasons for this inefficiency are unclear, it is unlikely that gene delivery is responsible given the high efficiency of retrovirus transduction (>90%). Chemicals that relax chromatin, such as inhibitors of histone deacetylation and DNA methylation, can increase the efficiency of iPSC creation [3, 4], suggesting that chromatin structure in host cells may limit nuclear reprogramming through poorly understood epigenetic mechanisms.
Oct4, Sox2, and Nanog constitute a central triad in the molecular circuitry underlying the maintenance of the undifferentiated state and pluripotency of ESCs [5, 6]. These three transcription factors activate their own genes and other pluripotency genes, in addition to suppressing aberrant expression of lineage-specific genes. The Oct4 gene contains binding sites for the Oct4 and Sox2 proteins in its distal enhancer region . However, transduction of the Oct4 gene alone or in combination with Sox2 and Nanog can not effectively activate Oct4 gene expression.
In contrast to Oct4, MyoD, a master transcription factor for skeletal myogenesis, can activate many of its target genes in nonmuscle cells, including its own gene and other skeletal muscle-specific genes, without the aid of additional exogenous factors [7–10]. Because of this potent transcriptional activation (transactivation), MyoD can directly convert one cell type into another, as exemplified by the creation of myotubes from pigmented retinal epithelial cells approximately 10 days after transduction of the MyoD gene . These findings suggest that MyoD can recruit various transcription factors and chromatin remodeling proteins to its target genes more efficiently than Oct4, leading to activation of suppressed genes embedded in closed chromatin. Therefore, we hypothesized that fusion of a MyoD fragment to full-length Oct4 might confer upon Oct4, the powerful transactivation capacity of MyoD, without losing the target specificity of Oct4. Through testing this hypothesis, we established a novel approach for dramatically increasing chromatin accessibility and recruitment of chromatin remodeling proteins to the Oct4 site, resulting in radical acceleration of iPSC generation.
MATERIALS AND METHODS
Preparation of Mouse iPSCs
Full-length and deletion mutants of mouse Oct4 cDNA were fused with various transactivation domains (TADs) and inserted into the pMXs-IP vector . Details of the subcloning procedures for each construct are available upon request. Polycistronic cDNAs encoding Sox2, Klf4, and c-Myc were transferred from the 4F2A lentiviral vector  to the pMXs-IP vector. pMXs-IP vectors encoding OSKM separately (Addgene, Cambridge, MA, www.addgene.org) were also used in some experiments. These pMXs-IP vectors were transfected into Plat-E cells  with Fugene six (Roche, Penzberg, Germany, www.roche-applied-science.com). Virus supernatant was harvested 48- and 72-hour later and filtered through a 0.45-μm syringe filter. MEFs were prepared from Oct4-green fluorescence protein (GFP) mice which harbor an internal ribosome entry site–GFP fusion cassette downstream of the stop codon of the Oct4 gene (Jackson Laboratory [Bar Harbor, ME, www.jax.org] #008214) . All animal experiments were conducted in accordance with the animal experiment guidelines of University of Minnesota. For chimera experiments, MEFs were prepared from mice that harbor the Oct4-GFP allele and ROSA26-lacZ allele. MEFs were seeded at 3 × 105 cells per 6-cm-dish on day −2 in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS). Fresh virus supernatant was added to MEFs on days −1 and 0 with 10 μg/ml polybrene. Culture medium was then changed to iPSC medium (DMEM, 15% FBS, 100 μM minimum essential medium [MEM] nonessential amino acids, 55 μM 2-mercaptoethanol, 2 mM L-glutamine, and 1,000 u/ml leukemia inhibitory factor) on day 1. Transduced MEFs were subcultured onto irradiated SNL feeder cells at 2 × 105cells per 6 cm dish on day 4 and maintained on the feeder cells in protocol A. The maximum number of GFP-positive colonies obtained around day 18 was divided by 2 × 105 to obtain the efficiency of making iPSCs. In protocol B, transduced MEFs were maintained without feeder cells. GFP-positive colonies were picked up around day 10 to clone without feeder cells for pluripotency analyses. Retrovirus titer was measured using NIH3T3 cells as described . All recombinant DNA research was conducted following the NIH guidelines
Preparation of Human iPSCs
Full-length human OCT4 cDNA fused with the M3 domain of human MYOD at the amino terminus was inserted into the pMXs-IP vector. pMXs-IP vectors encoding human M3O and OSKM (Addgene) were transfected into Plat-A cells (Cell Biolabs, San Diego, CA, www.cellbiolabs.com) with Lipofectamin 2000 (Invitrogen, Carlsbad, CA, www.invitrogen.com). Virus supernatant was harvested 48 and 72 hours later (days −1 and 0, respectively, below), filtered through a 0.45-μm syringe filter, and transduced into dermal fibroblasts obtained from a 34-year-old Caucasian female (Cell Applications, San Diego, CA, www.cellapplications.com). On day −2, 2.7 × 104 fibroblasts were plated in each well of a 12-well plate in DMEM with 10% FBS. Fresh virus supernatant was added to the fibroblasts on days −1 and 0 with 10 μg/ml polybrene. On day 3, cells were harvested with trypsin and subcultured at 1.7 × 104 cells per well in 12-well plates coated with BD Matrigel hESC-qualified Matrix (BD Biosciences, San Jose, CA, www.bdbiosciences.com) in human iPSC medium (KnockOut DMEM/F-12 [Invitrogen], 20% Knockout Serum Replacement [Invitrogen], 100 μM MEM nonessential amino acids, 1% insulin-transferrin-selenium [Invitrogen], 0.1 mM 2-mercaptoethanol, 2 mM L-glutamine, and 4 ng/ml basic FGF). The medium was changed every other day.
Chromatin Accessibility to NsiI
One million cells were resuspended in ice-cold lysis buffer containing 0.1% NP40 and incubated on ice for 5 minutes as described previously . Nuclei were isolated with centrifugation at 4,000g for 5 minutes and digested with 200 u/ml NsiI for 2 hours at 37°C. DNA was purified and double-digested with MspI and BamHI, followed by Southern blotting using the radioactive probe shown in Figure 5D.
Additional methods are described in Supporting Information Materials and Methods section.
Generation of Mouse iPSCs with Heterologous TADs
We tested our above-mentioned hypothesis with a series of chimeric proteins, in which full-length mouse Oct4 was fused with various fragments of mouse MyoD (Fig. 1A). The basic helix-loop-helix domain of MyoD, important for dimerization and DNA binding, was not included in these constructs to avoid activation of MyoD-target genes. Each chimeric gene was cotransduced with a polycistronic retroviral vector encoding mouse Sox2, Klf4, and c-Myc (SKM)  into MEFs derived from Oct4-GFP mice, which contain the GFP gene knocked into the Oct4 locus . In this model, formation of GFP-positive colonies indicates that individual MEFs develop into Oct4-expressing cells capable of clonal growth. Expression of chimeric proteins was confirmed through immunoblotting with antibodies against Oct4 (Supporting Information Fig. S1). As a control, we transduced MEFs with OSKM (O-MEFs) on days −1 and 0 and transferred these cells onto SNL feeder cells on day 4 following a standard protocol (Fig. 1B, protocol A). GFP-positive colonies emerged around day 10, gradually increasing in number until reaching a peak by day 18. To calculate the percentage of MEFs that were reprogrammed into iPSCs, we divided the number of GFP-positive colonies by the total number of MEFs seeded in a culture dish. We estimated that 0.08% ± 0.09% of O-MEFs were converted into GFP-positive cells, which is similar to previous reports [2, 3] (Fig. 1A, right column). We then transduced MEFs with each chimeric gene along with SKM and followed the protocol described above (protocol A). M3O with SKM (M3O-SKM) increased the percentage of GFP-positive colonies most drastically, with 5.10% ± 0.85% of MEFs (M3O-MEFs) being transformed into GFP-positive cells by day 15. The M3 region encompasses the acidic TAD of MyoD (amino acids 3–56) , and acidity is considered important to its transactivation capability. However, the simple presence of acidity was insufficient to facilitate iPSC formation, as evidenced by a lack of increase in GFP-positive colonies in MEFs transduced with M6O, which also contains the main acidic amino acid cluster, or a chain of 20 glutamic acids attached to Oct4 (EO) (Fig. 1A). The high efficiency with which M3O created iPSCs when compared with Oct4 was not simply due to a difference in the retrovirus titer for the two virus suspensions. The titer for the M3O virus and Oct4 virus was 1.8 ± 0.2 × 107 and 2.1 ± 0.4 × 107 colony forming units per milliliter, respectively.
While conducting the above experiments, we noticed that GFP-positive colonies emerged from M3O-MEFs on approximately day 5 without transfer onto feeder cells (Fig. 1B, protocol B), and these colonies steadily increased in size and number (Fig. 1C). By approximately day 12, 3.6% ± 0.5% of M3O-MEFs formed GFP-positive colonies in the absence of feeder cells, probably supported by the surrounding MEFs serving as “autologous” feeder cells (Fig. 1D). In contrast, GFP-positive colonies emerged from O-MEFs between days 16 and 18 at an extremely low efficiency (0.0035% ± 0.0006%) with the same protocol. We then tested if GFP-positive colonies could be obtained without Sox2, Klf4, or c-Myc in the presence of M3O with protocol B (Fig. 1D). Although M3O still required Sox2 and Klf4, c-Myc was dispensable. Previous studies have reported that iPSCs can be established without c-Myc [18, 19]; however, the uniqueness of M3O-SK lies in the speed and efficiency with which GFP-positive colonies form. Although it requires 3–4 weeks and the presence of feeder cells for OSK to induce GFP-positive colonies at an efficiency of around 0.01% [18, 19], M3O-SK could generate GFP-positive colonies without feeder cells by day 7 after transduction at an efficiency of 0.44%, over 40-fold more efficient than OSK.
These striking differences between M3O and Oct4 prompted us to evaluate the specificity of the M3O configuration in relation to other host factors and TADs taken from other transcription factors using protocol B. First, we changed the location and number of the M3 domain in the fusion protein with Oct4 (Fig. 1E). Second, the two TADs in Oct4  were replaced with the M3 domain in various combinations (Fig. 1F). Third, the M3 domain was fused to Sox2 or Klf4 and tested in combination with other members of OSKM and M3O (Fig. 1G). Although OM3 was as effective as M3O in iPSC creation, all other constructs were virtually ineffective. In fourth experiment, TADs taken from other powerful transactivators were fused to Oct4 (Fig. 1H), including the TADs from Tax of human T-lymphotropic virus type-1 , Tat of human immunodeficiency virus type-1 [22, 23], Gata4 [24, 25] and Mef2c . When these TADs were fused to Oct4, none of them formed iPSC colonies. In conclusion, these TAD studies demonstrate that the M3 domain was unique among the tested TADs in that it dramatically improved the efficiency of making GFP-positive colonies when fused to Oct4. Therefore, we focused on M3O in the following studies.
Characterization of M3O-iPSCs
The GFP-positive colonies that emerged on day 5 following transduction with M3O-SKM using protocol B contained 31–143 cells in 12 colonies, with a median of 43 cells per colony. This number of cells would be produced after less than seven cell divisions assuming even division for each cell, which is strikingly small when compared with the median of 70 cell divisions needed before GFP-positive cells appear with OSKM . The colonies that emerged with M3O-SKM were usually homogenously GFP-positive from the beginning. On day 7, approximately 97% of these colonies were homogeneously GFP-positive with protocol B when compared with approximately 5% of colonies derived with OSKM obtained on day 12 with protocol A (Fig. 2A). Protocol A was used for OSKM because protocol B failed to yield a sufficient number of GFP-positive colonies. As a result, GFP-positive colonies were harvested at different time points corresponding to 2 days after the onset of GFP activation.
We further compared the quality of GFP-positive colonies obtained with M3O-SKM and OSKM by quantitative reverse-transcription polymerase chain reaction (qRT-PCR) analysis of three pluripotency genes (endogenous Oct4, endogenous Sox2, and Nanog) and three fibroblast-enriched genes (Thy1, Col6a2, and Fgf7) [27–29]. We selected homogeneously GFP-positive colonies obtained with M3O-SKM using protocol B and those with OSKM using protocol A to represent the best colonies for each group. Although cells were harvested at different time points corresponding to the onset of GFP activation, the interval between time points is the same. For OSKM, expression of the three pluripotency genes gradually increased during the initial week after emergence of GFP-positive colonies, indicating a slow maturation process toward pluripotency (Fig. 2B). For M3O-SKM, in contrast, levels of these transcripts reached or exceeded those seen in ESCs at the time of the emergence of GFP-positive colonies and remained at similar levels until day 30. This differential efficiency of transcriptional reprogramming was also evident with suppression of the three fibroblast-enriched genes. For M3O-SKM, expression levels of these genes on day 5 when the GFP signal was apparent were comparable to those seen in ESCs, but it took around 1 week after the activation of GFP for OSKM to accomplish the same level of gene suppression (Fig. 2C). Together, these results indicate that M3O-SKM can reprogram MEFs to an iPSC state more efficiently than OSKM.
The pluripotency of iPSC clones prepared with M3O-SKM following protocol B (M3O-iPSCs) was verified using three standard approaches. First, genome-wide transcript analysis demonstrated highly similar gene expression in M3O-iPSCs and ESCs. Out of 41,160 probes, 3,293 were greater than fourfold differentially expressed (upregulated or downregulated) in both ESCs and cloned iPSCs when compared with MEFs (Fig. 3A). The commonly upregulated genes included eight ESC-enriched genes, such as Oct4, Sox2, and Nanog. In addition, Thy1, Col6a2, and Fgf7 were downregulated more than 16-fold in both ESCs and iPSCs. Second, intramuscular injection of M3O-iPSCs into an nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mouse resulted in teratoma formation as shown by the presence of various tissues derived from the three germ layers (Fig. 3B). Third, aggregation of eight-cell stage embryos of the ICR strain with M3O-iPSCs containing the Oct4-GFP allele and ROSA26-lacZ allele formed chimeric mice (Fig. 3C, 3D). M3O-iPSCs contributed to germ cells in some chimeric mice (Fig. 3E). When one of the chimeric males (Fig. 3D, left) was crossed with a wild-type female ICR mouse (Fig. 3F, white adult at bottom), all 11 pups showed agouti or black coat color (Fig. 3F).
Establishment of Human iPSCs with M3O-SKM
We next evaluated if M3O could also facilitate generation of human iPSCs in comparison with OSKM. We transduced human M3O-SKM and OSKM into human dermal fibroblasts prepared from a 34-year-old female. Because these cells did not harbor a transgene that could be used as a convenient marker for reprogramming, expression of the pluripotency protein NANOG was monitored by immunofluorescence staining as an iPSC indicator. NANOG-positive human ESC-like colonies emerged around day 8 with M3O-SKM, with the number increasing by around day 15 when 0.30% ± 0.033% of the original fibroblasts were converted to iPSC colonies (Fig. 4A, 4B). In contrast, when OSKM was transduced, NANOG-positive colonies did not emerge until around day 12 and eventually only 0.0052% ± 0.0018% of the fibroblasts were turned into iPSC colonies. This indicates 58-fold increased efficiency with M3O-SKM in comparison with OSKM. Furthermore, while less than 10% of the colonies that appeared with OSKM were NANOG positive, more than 90% of the colonies produced with M3O-SKM were NANOG-positive, consistent with the results for mouse iPSCs. Cloned iPSCs prepared with M3O-SKM also expressed endogenous OCT4 and surface markers stage-specific embryonic antigen 4 (SSEA4), TRA-1-60, and TRA-1-81 on day 28 (Fig. 4C). Transduced M3O was suppressed by this day (not shown). In addition, iPSCs prepared with M3O-SKM expressed 12 pluripotency genes as demonstrated by qRT-PCR (Fig. 4D). All the 20 mitotic spreads prepared from a cloned M3O-SKM iPSCs demonstrated normal karyotypes (Fig. 4E). Finally, they formed teratomas when injected into an NOD/SCID mouse (Fig. 4F), proving pluripotency of the cells.
Chromatin Analyses of Pluripotency Genes in M3O-MEFs
We used only mouse cells in all of the following experiments. To understand how M3O-SKM facilitated nuclear reprogramming at the molecular level, we examined several chromatin changes at the Oct4 gene during the early phase of iPSC generation. All analyses were performed with protocol B on all MEFs in a culture dish including GFP-positive and GFP-negative cells without subculture for 9 days. First, we studied changes in DNA methylation at the promoter of the Oct4 gene. CpG dinucleotides at the proximal promoter of the Oct4 gene are heavily methylated in MEFs, unlike in ESCs and iPSCs  (Fig. 5A), and this serves as a major inhibitory mechanism for Oct4 transcription. Although the number of unmethylated CpG sites remained essentially the same on day 9 in O-MEFs, the number increased approximately twofold in M3O-MEFs on the same day (Fig. 5A, 25.5% vs. 55.5%). The more advanced demethylation in M3O-MEFs than in O-MEFs is consistent with the higher percentage of GFP-positive cells in M3O-MEFs than in O-MEFs on day 9 (12.77% vs. 0.52%) as shown by flow cytometry (Fig. 5B).
Next, we studied the binding of Oct4 and Sox2 to the distal enhancer of the Oct4 gene  using chromatin immunoprecipitation (ChIP). The binding of Oct4 and Sox2 to the distal enhancer remained low with O-MEFs (Fig. 5C). Importantly, however, Oct4, which was identical to M3O in this case, was already highly bound to the Oct4 distal enhancer in M3O-MEFs as early as day 3 when no GFP-positive colonies had appeared yet (Fig. 5C, the red column in the Oct4 panel). The Oct4-binding level gradually increased subsequently, eventually reaching the level comparable to that seen in ESCs on day 9. The chromatin binding of Sox2 displayed a similar tendency. It was important to understand if the high binding levels of Oct4 and Sox2 on day 9 could be explained by the existing GFP-positive cells (12.77%, Fig. 5B). However, it was technically difficult to collect GFP-negative cells in the order of 1 × 107 cells with a fluorescence activated cell sorter (FACS) for ChIP analyses. As an alternative approach, we examined the binding levels of these two proteins in the mixture of ESCs and MEFs at a 13:87 ratio. This study showed substantially lower binding of Oct4 and Sox2 in comparison with the day 9 levels in M3O-MEFs (Fig. 5C, ESCs+MEFs in blue). This observation indicates that Oct4 and Sox2 were bound to the Oct4 enhancer in the majority of M3O-MEFs including GFP-negative cells on day 9. The increased binding of these two proteins to chromatin in M3O-MEFs prompted us to investigate if chromatin accessibility at the distal enhancer was also increased in M3O-MEFs. Increased chromatin accessibility is generally indicated by higher sensitivity to DNAses . We digested chromatin from M3O-MEFs and O-MEFs with the restriction enzyme NsiI and analyzed DNA fragments using Southern blotting. Indeed, chromatin accessibility was consistently higher in M3O-MEFs when compared with O-MEFs between days 5 and 9 (Fig. 5D). Additionally, we selected GFP-negative M3O-MEFs with a FACS on day 9 followed by NsiI digestion analysis. This GFP-negative population also demonstrated increased sensitivity to NsiI (Fig. 5D, far right), indicating that the minor GFP-positive population did not significantly influence the overall result of chromatin accessibility.
Previous reports have shown that the Paf1 complex is recruited to the distal enhancer of the Oct4 gene through binding to the Oct4 protein [33, 34] and then generally moves to the coding region of the gene . We observed that three Paf1 complex subunits—parafibromin, Leo1, and Paf1—displayed a gradual increase of binding to the distal enhancer and coding region of the Oct4 gene in M3O-MEFs, but not in O-MEFs, between days 3 and 9 following transduction (Fig. 5C). The Paf1 complex recruits the histone methyltransferase complex COMPASS, which catalyzes trimethylation of lysine four on histone H3 (H3K4me3) . This histone modification, a marker for active genes, was also increased specifically in M3O-MEFs in the coding region of the Oct4 gene (Fig. 5E). Two other markers for active genes, acetylation of lysines 9 and 14 on histone H3 (H3K9ac and H3K14ac) , were also increased specifically in M3O-MEFs (Fig. 5E). In addition, two markers for suppressed genes, trimethylation of H3K9 (H3K9me3) and H3K27 (H3K27me3) , were more drastically decreased in M3O-MEFs than those in O-MEFs (Fig. 5F). Similar results were observed at the Sox2 locus (Supporting Information Fig. S2). Among these chromatin changes, the levels of H3K9me3 and H3K27me3 in M3O-MEFs most quickly reached the levels observed in ESCs (Fig. 5F), suggesting that the loss of these suppressive histone markers precedes other chromatin modifications. Taken together, these results demonstrate that chromatin at Oct4 and Sox2 loci was generally remodeled in majority of M3O-MEFs, including the GFP-negative population, toward an ESC pattern during the first 10 days after transduction, while chromatin in the majority of O-MEFs was not significantly altered.
In addition to global chromatin remodeling, M3O-SKM also elicited wider expression of two pluripotency markers than OSKM: alkaline phosphatase and SSEA1. Alkaline phosphatase was almost ubiquitously expressed by day 9 in M3O-MEFs, unlike the weak and heterogeneous expression observed in O-MEFs (Fig. 6A). SSEA1 was also more widely expressed in M3O-MEFs than in O-MEFs by day 9 as shown by immunofluorescence microscopy and flow cytometory (Fig. 6B, 6C). While alkaline phosphatase and SSEA1 are not exclusively expressed in pluripotent cells, these findings support the above interpretation that M3O-SKM remodeled the chromatin in much more wider population of the cells to a certain degree unlike OSKM. Rapid cell proliferation is known to facilitate iPSC generation as shown using p53-null MEFs ; however, neither M3O-SKM nor M3O alone facilitated MEF proliferation during the initial 9 days after transduction (Fig. 6D, 6E).
Chromatin Analyses of Pluripotency Genes Without c-Myc
M3O-SK induced GFP-positive colonies over 100-fold more efficiently than OSKM with protocol B (0.44% with M3O-SK in Fig. 1D vs. 0.0035% with OSKM in Fig. 1F). This observation suggests that the M3 domain could compensate for the lack of c-Myc when Oct4 activation was used as an indicator. Although several roles of c-Myc have been proposed, its precise function in iPSC formation remains elusive . To further understand the roles of c-Myc in the activation of pluripotency genes, we repeated chromatin analyses at the Oct4 and Sox2 loci comparing MEFs transduced with three genes (M3O-SK or OSK) and four genes (M3O-SKM of OSKM) on day 9 when the effects of M3O-SKM were readily detectable. We also transduced one gene (M3O or Oct4) and two genes (M3O+Sox2 or Oct4+Sox2) for comparison. At this time point, 3.16% of MEFs were GFP-positive with M3O-SK (Fig. 7A), and almost no GFP-positive cells were observed with other combinations of one, two, or three genes. However, M3O-SK did not significantly decrease the overall level of DNA methylation when compared with other gene combinations (Fig. 7B).
As for transcription factor binding to the enhancer, M3O facilitated binding of Oct4, Sox2, and parafibromin in combination with Sox2 or Sox2 and Klf4 (Fig. 7C, red), with some of these binding levels comparable to levels achieved with M3O-SKM. However, Leo1 and Paf1 were not recruited to the enhancer without c-Myc (Fig. 7C). The binding of parafibromin, Leo1, and Paf1 to the initiation site of Oct4 was also weak without c-Myc (Supporting Information Fig. S3A). Consistent with this partial assembly of the Paf1 complex at the Oct4 gene, the level of H3K4me3 remained low without c-Myc (Fig. 7D; Supporting Information Fig. S3B). Another active gene maker, H3K9ac, also remained low without c-Myc (Fig. 7D; Supporting Information S3B). Whereas H3K9me3 was effectively decreased by M3O-S and M3O-SK, H3K27me3 was more resistant to demethylation by any of the gene combinations without c-Myc (Fig. 7E). At the Sox2 gene, when compared with the Oct4 gene, M3O did not substantially increase the binding of Oct4 or Sox2 to the enhancer alone or in combination with Sox2 or Sox2 and Klf4 (Supporting Information Fig. S4A). The changes in the levels of H3K4me3, H3K9ac, H3K9me3, and H3K27me3 were all weak in the absence of c-Myc (Supporting Information Fig. S4B). Together, these chromatin studies indicate that while M3O could facilitate formation of GFP-positive colonies without c-Myc, the overall level of chromatin remodeling in GFP-negative MEFs was low in the absence of c-Myc.
This study advances the field of iPSC biology by showing that one of the rate-limiting steps in iPSC formation with OSKM is poor chromatin accessibility at pluripotency genes and that a strong TAD can overcome this problem. Because iPSC formation was dramatically improved with M3O-SKM, the factors required to increase chromatin accessibility most likely already exist within MEFs but are not effectively recruited to pluripotency genes when using OSKM. Our current working model is that the MyoD TAD overcomes the barrier of closed chromatin by effectively attracting chromatin modifying proteins and thereby facilitating the binding of Oct4 and other regulatory proteins as well as epigenetic modifications at pluripotency genes (Fig. 7F). Myc family proteins have been proposed to globally relax chromatin in part through activation of the histone acetyltransferase GCN5 and in part through direct binding to thousands of genomic loci [37, 39]. Our results also support c-Myc's potential roles in chromatin remodeling although we do not know if c-Myc is located upstream of chromatin remodeling, as suggested by the direct binding of c-Myc to pluripotency genes , or downstream of chromatin remodeling (preferential proliferation of MEFs with remodeled chromatin).
One of the central questions related to the molecular mechanisms of iPSC formation is how closed chromatin at the loci of Oct4, Sox2, and Nanog are opened by OSKM. Little is known about this mechanism. One potential mechanism is that chromatin disruption occurs during repeated DNA replication as suggested by a report that 92% of B lymphocytes derived from inducible OSKM transgenic mice become iPSCs after 18 weeks of culture . Additionally, knockdown of p53 in B cells shortened both cell doubling time and the time required to form iPSCs by twofold. However, this does not seem to be the case for M3O-SKM because it did not facilitate cell proliferation. Additionally, emerging GFP-positive colonies contained far less cells than their counterparts obtained from B cells. It has been difficult, if not impossible, to perform biochemical analysis of the early process of iPSC formation, such as epigenetic remodeling at pluripotency genes, because of the presence of feeder cells and nonresponsive MEFs that comprise more than 90% of transduced cells. However, the MyoD TAD eliminated the requirement for feeder cells and achieved significant levels of epigenetic remodeling even in those MEFs that eventually fell short of activating GFP with protocol B. Thus, the MyoD TAD is expected to facilitate the dissection of epigenetic processes during the early phase of iPSC formation.
Wang et al.  recently demonstrated that the TAD of the herpes simplex virus type 1 protein VP16 facilitates iPSC generation when fused to Oct4, Sox2, and Nanog. The efficiency of making iPSCs was similar between M3O and Oct4-VP16 when combined with Sox2 and Klf4 (0.44% with M3O in the absence of feeder cells in Fig. 1D vs. 0.47% with Oct4-VP16 in the presence of feeder cells). However, the functions of the M3 domain and the VP16 TAD are different in three points. First, germline-competent iPSCs can be established with Oct4-VP16 without additional factors but M3O alone did not generate iPSCs. Second, while the M3 domain was not effective in iPSC formation when fused to Sox2 or Klf4, the VP16 TAD was also effective when fused to Sox2 and Klf4 although the efficiency drops to approximately 0.2%. Third, whereas three tandemly repeated VP16 TADs were more powerful than a single VP16 TAD, two tandemly repeated M3 domains were inhibitory. These differences suggest that M3O and Oct4-VP16 promote iPSC production based on different mechanism. As for the binding proteins, many chromatin proteins are known to bind to the VP16 TAD (see  for references), but only p300/cAMP response element-binding (CREB)–binding protein is known to interact with the MyoD TAD . Related to this, epigenetic remodeling in the MEFs transduced with Oct4-VP16 has not been characterized. The fact that these two independent studies achieved similar results of increasing iPSC production validates the power of the TAD-fusion approach in nuclear reprogramming.
By combining transcription factors with TADs, this approach to nuclear reprogramming is expected to have a range of applications from inducing pluripotency, as shown in this study, to inducing direct conversion from one differentiated cell type to another without transitioning through iPSCs [25, 43, 44]. As pointed out before, the first iPSC article initiated a new approach of identifying an essential combination of genes for nuclear reprogramming from a pool of candidate genes . First, these authors simultaneously transduced all candidate genes into MEFs and, after detecting a positive effect, systematically eliminated each gene to identify a necessary and sufficient set of genes. The power of this simple and methodical approach has been repeatedly demonstrated by successful application of the same approach to identifying different sets of three sufficient genes in making pancreatic β cells, neurons, and cardiac muscle cells from other cell types by direct conversion [25, 43, 44]. However, the chance of success with this approach is constrained by an inherent limitation of the transactivation capacity of each gene product. The strategy of TAD-fusion to potentiate transactivators is expected to break this limitation and further advance the study of nuclear reprogramming. Although only the MyoD TAD was effective in our small-scale screen of TADs, the effect of each TAD is highly likely to be dependent on cell types, host transcription factors, and target genes. Other TADs have been used to amplify the activity of transcription factors. For instance, the TAD of VP16 was fused to the transcription factor Pdx1 to facilitate conversion of hepatocytes to pancreatic cells [38, 45]. However, the MyoD TAD has not been used in nuclear reprogramming. In principle, a TAD-fusion approach could be applied to combinations of many other transcription factors and TADs to amplify the activity of the host transcription factor and control cell fate decisions.
We demonstrated that the fusion gene between the MyoD TAD and Oct4 facilitated iPSC formation in the absence of feeder cells and c-Myc. The TAD promoted epigenetic remodeling at pluripotency genes during iPSC formation even in the cells that failed to be completely reprogrammed. Our results indicate that poor chromatin accessibility and factor recruitment is a major limiting factor of iPSC creation.
We thank Susan Keirstead and Michael Franklin for critical reading of the manuscript; Atsushi Asakura for the support of mouse experiments; James Dutton, Lucas Greder, Kimberly Stelzig, Kristin Voltzke, and Liying Zhang for technical support for iPSC culture; Timothy D. O'Brien for histological analyses; Wei-Shou Hu, Kartik Subramanian, and Wayne Xu for DNA microarray analyses; and Toshio Kitamura for Plat-E cells and pMXs-IP retroviral vectors. Karyotyping and histological preparation were performed at the University of Minnesota Cytogenetics Core Laboratory and the Comparative Pathology Shared Resource, respectively, in support with the comprehensive Masonic Cancer Center NIH grant (P30 CA077598-09). This work was also supported by the Richard M. Schulze Family Foundation (to M.F. and N.K.), NIH (R01 DK082430), Leukemia Research Fund, and Academic Health Center of the University of Minnesota (to N.K.).
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.