Search tips
Search criteria 


Cell Rep. 2016 June 14; 15(11): 2550–2562.
Published online 2016 June 2. doi:  10.1016/j.celrep.2016.05.027
PMCID: PMC4914771

Defining the Minimal Factors Required for Erythropoiesis through Direct Lineage Conversion


Erythroid cell commitment and differentiation proceed through activation of a lineage-restricted transcriptional network orchestrated by a group of well characterized genes. However, the minimal set of factors necessary for instructing red blood cell (RBC) development remains undefined. We employed a screen for transcription factors allowing direct lineage reprograming from fibroblasts to induced erythroid progenitors/precursors (iEPs). We show that Gata1, Tal1, Lmo2, and c-Myc (GTLM) can rapidly convert murine and human fibroblasts directly to iEPs. The transcriptional signature of murine iEPs resembled mainly that of primitive erythroid progenitors in the yolk sac, whereas addition of Klf1 or Myb to the GTLM cocktail resulted in iEPs with a more adult-type globin expression pattern. Our results demonstrate that direct lineage conversion is a suitable platform for defining and studying the core factors inducing the different waves of erythroid development.

Graphical Abstract

An external file that holds a picture, illustration, etc.
Object name is fx1.jpg


Although several factors are known to participate in the conserved genetic program instructing development of committed erythroid progenitors, the minimal combination of factors required for direct induction of erythroid cell fate remains unknown. The identification of the key players controlling red blood cell (RBC) development is important for understanding basic biology and can be used to study and recapitulate erythropoiesis in vitro as well as to model and develop new therapies for RBC disorders (Tsiftsoglou et al., 2009). Fate decisions in erythropoiesis have been investigated extensively, focusing on lineage-specific transcription factors and cofactors as the main drivers of the process (Cantor and Orkin, 2002, Shivdasani and Orkin, 1996). Genes found to be essential for normal RBC development in mice include Gata1, Zfpm1, Lmo2, Klf1, Myb, Tal1, Runx1, and Ldb1 (Mead et al., 2001, Palis, 2014). However, the factors constituting the core transcriptional machinery that initiates and specifies erythroid cell fate are still unknown.

A major obstacle for defining core transcriptional networks is the difficulty of discriminating “instructive” factors from “permissive” factors. Numerous studies have demonstrated that it is possible to directly convert a mature cell type into another, bypassing the pluripotent state, using a defined set of lineage-instructive transcription factors (Jopling et al., 2011, Takahashi, 2012). This approach, called direct lineage reprogramming, can yield a wide range of clinically relevant cell types, such as neurons, cardiomyocytes, and hepatocytes (Huang et al., 2011, Ieda et al., 2010, Sekiya and Suzuki, 2011, Vierbuchen et al., 2010). Because the converted cells resemble their bona fide counterparts in terms of phenotype and function, direct lineage reprogramming is currently a widely investigated approach for generating defined cell types for regenerative medicine. In contrast to loss-of-function studies, direct reprogramming distinguishes absolutely essential cell fate-inducing factors from merely permissive factors, revealing the master regulators of specific cell lineages (Vierbuchen and Wernig, 2011). Therefore, we reasoned that direct lineage reprogramming is an unambiguous method for defining the core transcriptional machinery directing RBC development.

Several laboratories have described methods for reprogramming differentiated somatic cells to hematopoietic progenitors with multilineage potential (Batta et al., 2014, Pereira et al., 2013, Riddell et al., 2014, Szabo et al., 2010), whereas others have reported protocols of direct induction to the erythroid lineage starting from B cells (Sadahira et al., 2012) and pluripotent cell sources (Weng and Sheng, 2014). However, none of these studies have shown robust erythroid-restricted fate conversion from non-hematopoietic differentiated somatic cells.

Here we identify the transcription factors Gata1, Tal1, Lmo2, and c-Myc (GTLM) as the minimal set of factors for direct conversion of mouse and human fibroblasts into erythroid progenitors. The resulting cells, which we term induced erythroid progenitors/precursors (iEPs), resemble bona fide erythroid cells in terms of morphology, colony-forming capacity, and gene expression. While murine GTLM iEPs express both embryonic and adult globin genes, the addition of Klf1 or Myb induces a switch in globin gene expression to generate iEPs with a predominant definitive-type globin expression pattern. This approach can be used as a model for understanding, controlling, and recapitulating erythroid lineage development and disease.


A Combination of Transcription Factors Induces the Erythroid Fate in Murine Fibroblasts

We hypothesized that overexpression of transcription factors involved in hematopoietic and, specifically, erythroid development in fibroblasts could directly convert these cells into erythroid progenitors or precursors. A retroviral library was created from mouse fetal liver (FL) cDNA expressing the coding region of 63 candidate factors (Table S1). Adult tail tip fibroblasts (TTFs) were derived from erythroid lineage-tracing mice (Heinrich et al., 2004), which express the yellow fluorescent protein (eYFP) from the Rosa26 locus in all cells that have expressed the erythropoietin receptor (Epor, Cre knocked into one allele of the endogenous Epor locus) transcript at any stage of their development (Figure 1A). In vivo, the expression of eYFP is first detected in bipotent progenitors of megakaryocytes and erythrocytes (pre-MegEs) and is subsequently robustly expressed in erythroid progenitors (Singbrant et al., 2011). Importantly, eYFP was never detected in other hematopoietic lineages or cell types examined. TTF cultures were carefully depleted of hematopoietic cells by magnetic separation using a cocktail of nine hematopoietic antibodies (Experimental Procedures) and passaged at least three times prior to transduction to obtain pure fibroblast cultures. The primary readout for erythroid lineage conversion was the formation of colonies of eYFP+ (EpoR+) round cells.

Figure 1
Forced Expression of Gata1, Tal1, Lmo2, and c-Myc Reprograms Murine Adult Fibroblasts into Erythroid Progenitors

Through screening multiple combinations of candidate factors, a combination of seven transcription factors was identified (Nfe2, Myb, Klf1, Gata1, Tal1, Lmo2, and c-Myc), that, in 8 days, converted TTFs into clusters of round cells displaying an erythroid precursor-like morphology. These clusters were eYFP+ (EpoR+), indicating induction of the erythroid transcriptional program (Figure 1B). Hence, we termed these cells iEPs.

Only Four Factors Are Necessary and Sufficient for iEP Generation

To identify the minimal set of factors required for iEP generation, we performed single factor subtraction experiments. We found that removal of Gata1, Tal1, Lmo2, or c-Myc from the factor cocktail completely abrogated iEP formation (Figure 1C; Figure S1). Notably, TTF reprogramming to iEPs was significantly enhanced using only these four factors compared with the initial seven factors (Figure 1C). Thus, we concluded that GTLM factors constitute the minimal set required for iEP generation.

iEPs Exhibit Morphological and Gene Expression Properties of Erythroid Cells

To better characterize iEP emergence, we examined reprogramming at different time points: on day 5, when the first YFP+ clusters of round cells appeared, and on day 8, when large YFP+ colonies could be observed (Figure 2A). On day 5, iEPs displayed an erythroid precursor-like morphology, featuring a characteristic central nucleus, coarse chromatin, and blue cytoplasm after May-Grünwald-Giemsa staining (Figure 2B). Some cells also stained weakly positive with benzidine stain and appeared mildly red when pelleted, indicating that the cells contained hemoglobin (Figures 2C and 2D). On day 5, a small fraction of iEPs co-expressed eYFP and the erythroid-specific surface marker Ter119 (Figure 2E). iEPs harvested on day 8 presented a more differentiated erythroid phenotype. They were significantly smaller, accumulated more hemoglobin, and upregulated Ter119 expression (Figures 2B–2F). Very few enucleated reticulocytes were observed, suggesting inefficient enucleation (Figure 2G). Kinetic analysis by flow cytometry revealed that erythroid precursor output was highest on day 6, followed by day 8, with 10.5% ± 4.6% and 6.6% ± 0.5% of live YFP+ cells co-expressing CD71 and Ter119, respectively (Figure S2A). Furthermore, the pan hematopoietic marker CD45, which is downregulated in erythroid cells, was not expressed at any time during reprogramming. This suggests that GTLM reprogramming is direct and does not involve an intermediate hematopoietic progenitor stage.

Figure 2
Induced Erythroid Progenitors Exhibit Properties of Bona Fide Erythroid Cells

As expected from the erythroid flow cytometric profile and cell morphology, day 8 iEPs downregulated the expression of fibroblast-specific genes and upregulated the expression of erythroid cell-specific genes (Figure 2H). We then analyzed the expression of the different globin genes from both the α and β globin clusters, which are differentially expressed throughout development. In the mouse, all globin genes of the α (Hba-x, Hba-a1, and Hba-a2) and the β globin loci (Hbb-y, Hbb-bh1, Hbb-b1, and Hbb-b2) are expressed in primitive erythroid cells, whereas only adult globin genes (Hba-a1, Hba-a2, Hbb-b1, and Hbb-b2) are expressed in definitive erythroid cells (Kingsley et al., 2006). Day 8 iEP expressed both embryonic and adult globins, with predominant expression of the embryonic types (Figure 2I). Specifically for the β-globin locus, iEPs expressed 50 times more embryonic Hbb-y than adult Hbb-b1, suggesting that, although adult fibroblasts were used, GTLM induces an erythroid program that is more similar to primitive than definitive erythropoiesis.

Next, we assayed the in vitro differentiation capacity of both day 5 and day 8 iEPs by colony-forming assays. After 8 days in methylcellulose supplemented with human Erythropoietin (hEPO), murine stem cell factor (mSCF), and dexamethasone, iEPs formed two types of colonies: distinctly red (red iEP) and not visibly red (non-red iEP) (Figure 2J). Although cells from red iEP colonies displayed erythroblast morphology, cells from non-red colonies were irregular, had a large deep blue and granular cytoplasm, and did not resemble erythroid cells (Figure 2J). Of the day 5 iEPs, approximately 1 in 1,000 formed red colonies, suggesting that most cells reprogrammed to a more differentiated erythroid cell state without colony-forming ability. This ratio was reduced in day 8 iEPs, which could be explained by iEPs undergoing differentiation from days 5–8 and/or the pMX vectors suffering silencing over time (Figure S2B). In parallel, we asked whether different stoichiometric ratios of GTLM factors could improve the iEPs’ erythroid colony output. Interestingly, day 5 iEPs generated with GTLM in a 2:1:1:1 ratio (D5-iEP+G) gave rise to red colonies almost exclusively, whereas day 5 iEPs generated with GTLM in a 1:1:1:2 ratio (D5-iEP+M) generated almost only non-red colonies (Figure 2K). Increasing the ratio of Tal1 or Lmo2 had no significant effect. These data imply that optimizing the stoichiometry of the reprogramming factors can further enhance the erythroid output.

In addition to adult tail tip fibroblasts, iEPs could also be generated from murine embryonic fibroblasts (Figures S2C–S2E), demonstrating that the GTLM factors can reprogram fibroblasts from other origins. GTLM induction of the erythroid fate is thus a rapid and direct process, yielding erythroblast-like cells with bona fide properties.

iEP-Derived Red Colonies Display a Gene Expression Signature Similar to Bona Fide Burst-Forming Unit-Erythroid Colony Cells

To characterize the reprogrammed cells at the molecular level, we performed global gene expression profiling comparing iEPs with bona fide erythroid progenitors and TTFs. To obtain RNA from pure cell populations, burst-forming unit-Erythroid (BFU-E) colony-forming assays were performed on day 5 iEPs as well as on mouse embryonic day (E) 14.5 FL and adult bone marrow (BM) cells. iEP-derived red and non-red colonies were picked separately, and untransduced TTFs were collected as controls. Unsupervised hierarchical clustering revealed that iEP-derived red colonies (red iEPs) clustered together with primary BFU-Es (FL colony and BM colony) (Figure 3A), indicating that their overall transcriptome is more similar to bona fide erythroid progenitors than to their starting fibroblast cell type. Genes differentially expressed more than 2-fold were mined for significantly overrepresented functional categories using the annotation tool DAVID (Database for Annotation, Visualization and Integrated Discovery) (Huang et al., 2009a, Huang et al., 2009b), demonstrating that genes induced in red iEPs were significantly associated with gene ontology (GO) terms relating to hematopoiesis, erythrocyte function, and development (Figure 3B). On the contrary, genes downregulated in red iEPs compared with starting fibroblasts were significantly associated with fibroblast-like function, such as “extracellular matrix organization,” consistent with the vast inactivation of the fibroblast gene expression program. Taken together, these data demonstrate that red iEP-derived colonies show large-scale downregulation of the fibroblast-specific program and extensive activation of genes specific to the erythroid lineage, indicating that GTLM factors are sufficient to trigger global transcriptional modeling toward the erythroid lineage.

Figure 3
iEP-Derived Red Colonies Retain a Primitive Erythroblast Expression Signature

iEP-Derived Non-red Colonies Show Incomplete Reprogramming

To characterize the iEP-derived non-red colonies, we analyzed the differentially expressed genes between red iEPs and non-red iEPs (Figure S3). We found that non-red iEPs lacked induction of genes associated with terminal erythropoiesis, such as globins and genes necessary for heme production (Figure S3A). Moreover, the expression levels of Gata1, Tal1, and Lmo2 were 5.6-, 5.2-, and 1.7-fold higher in red iEPs than in non-red iEPs, respectively (Figure S3B). This suggests that non-red iEP colonies have failed to completely reprogram and generate hemoglobinized cells, possibly as a result of inadequate GTL factor expression levels or factor stoichiometry.

The Gene Expression Profile of iEP-Derived Red Colonies Resembles that of Primitive Erythroblasts

Because red iEPs expressed mainly embryonic globin and did not enucleate efficiently, we asked whether their global gene expression profile resembled primitive rather than definitive erythroid cells. Interestingly, the embryonic globin genes (Hbb-y, Hba-x, and Hbb-bh1) were within the top seven differentially expressed genes between red iEPs and bona fide definitive FL and BM BFU-Es (Figure 3C), confirming that this distinctive feature of the primitive erythroid transcriptional program was active in iEPs compared with definitive erythroid cells. Next, we analyzed genes that were at least 4-fold more expressed in bona fide BFU-Es than in red iEPs (iEP/def-low) and genes that were at least 4-fold more expressed in red iEPs than in bona fide BFU-Es (iEP/def-high) and inspected their behavior in data within the Erythron database, which contains global gene expression profiles from primitive, fetal definitive, and adult definitive erythroid cells in mice (Kingsley et al., 2013). We found that iEP/def-low genes in general had lower expression in primitive erythroid cells from the yolk sac compared with the definitive erythroid cells from FL and BM, suggesting that iEPs retain an expression signature similar to that of primitive erythroblasts (Figure 3D). Regarding the iEP/def-high gene set, we did not observe a correlation with the yolk sac profile, although the three embryonic globin genes were among the top differentially expressed genes in this list (Figure 3E). This suggests that bulk iEPs primarily exhibit a molecular signature that is more comparable with a primitive than definitive erythroid program, possibly because additional factors are necessary to enable definitive erythropoiesis. These findings also raise the question of whether individual iEP clones are different so that some are close to primitive erythroid cells, whereas other clones are more similar to definitive erythroid cells.

Induction of Adult Globin Expression by Klf1 and Myb

We next asked whether the overexpression of additional factors could increase the expression of adult hemoglobin in iEPs at the expense of the embryonic globins. This would identify key regulators of hemoglobin expression in iEP as well as improve the potential of these cells for therapeutic applications, as highlighted by Trakarnsanga et al. (2014). We evaluated the transcription factors Sox6, Bcl11a, Klf1, and Myb, all previously identified to directly or indirectly downregulate the expression of embryonic and fetal globin genes (Sankaran et al., 2009, Sankaran et al., 2011, Yi et al., 2006, Zhou et al., 2010). When overexpressing the GTLM factors (4F) together with a single “switching” factor in wild-type (WT) TTF, only Klf1 and Myb increased the expression ratio of adult Hbb-b1 over embryonic Hbb-y compared with the 4F alone, although differences were not statistically significant (Figure 4A).

Figure 4
Klf1 and Myb Enhance Adult Hemoglobin Expression in Single iEPs

To investigate whether constitutive expression of switching factors could reinforce this trend, we cloned Klf1 and Myb independently into retroviral vectors including a blasticidin selection gene and generated TTF cell cultures with constitutive expression of Klf1 or Myb (Klf1 and Myb). iEP generated from these fibroblasts presented a reversed globin expression pattern and predominantly expressed adult Hbb-b1 globin (54.3% ± 13.9% for Klf1 and 61.3% ± 4.9% for Myb). This expression pattern is similar to that of Ter119+ cells harvested from E11 FL, where we still detected Hbb-y expression, possibly because of the presence of residual primitive erythroblasts (Figure 4A; McGrath et al., 2011). Thus, the persistence of Hbb-y in Klf1- and Myb-overexpressing cells could reflect the presence of clones with incomplete reprogramming to a definitive-type regulation of globin genes.

Next we investigated the clonal identity and heterogeneity of iEPs and tested whether the addition of Klf1 and Myb generated clones with a more definitive erythroid phenotype overall or merely affected globin expression. We sorted single YFP+ Ter119+ iEPs that were generated from either Epor reporter TTF or Epor reporter TTF with constitutive expression of Klf1 or Myb and performed single-cell qRT-PCR (Figure S4A). As for the globin genes, we ranked them from highest (R1) to lowest (R5) based on the inverse Ct value (50 – Ct = iCt) of each gene in each single cell (Figure S4B; Figure 4B). GTLM-iEPs expressed both embryonic and adult globins at the single-cell level, indicating that reprogramming generates iEPs with a mixture of globins (Figure 4B). Consistent with our previous observation, Klf1 and Myb overexpression increased the frequency of single cells with a “definitive” globin expression pattern. Of note, Klf1 affected the expression of both the α and β loci, whereas Myb only modified the β locus (Figure 4B). However, not all Klf1- and Myb-overexpressing cells displayed a definitive-type regulation of globin genes, denoting clonal heterogeneity in reprogramming and globin loci regulation.

Then we analyzed the expression of primitive-specific and definitive-specific genes in the three iEP subsets. We selected transcriptional regulators and other genes that were previously described to be differentially expressed between primitive and adult erythroblasts, regardless of their differentiation stage (see Table S5 in Kingsley et al., 2013). Unsupervised hierarchical clustering revealed that iEPs clustered together with primitive yolk sac erythroid cells (E9 YS), indicating that, based on the selected genes, iEPs are more similar to primitive than definitive erythroid cells (Figure 4C). The clustering did not change when globin and reprogramming genes were removed from the analysis (data not shown). Closer examination of gene groups revealed that single iEPs expressed some definitive-specific genes, such as Sox6 and Aldh1a1 (cluster 1), as well as some primitive-specific genes, such as Cited2 and Rragd (cluster 3). There was also a group of primitive-specific genes, such as Lin28b and Aqp3, whose expression was clearly suppressed in iEPs (cluster 2). Furthermore, the addition of Klf1 or Myb did not clearly change the expression of primitive- or definitive-specific genes compared with GTLM alone (Figure 4C). In conclusion, single-cell qRT-PCR demonstrates that iEP clones possess an expression profile reflective of a mixture between primitive and definitive erythropoiesis (Figure 4D) and that Klf1 and Myb do not induce a switch from a generally primitive to a definitive gene expression program but, instead, mainly act as globin switching factors in iEPs by increasing adult hemoglobin expression at the expense of embryonic globins.

Human iEPs Generated by Forced Expression of GTLM

To determine whether GTLM factors could induce the erythroid cell fate in the human setting, we transduced primary cultures of human foreskin fibroblasts (HFFs) from two independent sources with the murine versions of the four factors (Experimental Procedures), and evaluated erythroid reprogramming 12 days thereafter. GTLM-transduced HFFs gave rise to clusters of round cells that co-expressed CD71 and Glycophorin A (GPA), indicating the presence of early erythroid precursor-like cells (Figure 5A; Figure S5). 7.7% ± 5.9% of CD71+ GPA+ cells were also positive for Band 3, a membrane glycoprotein expressed in late erythroid precursors and mature RBCs (Chen et al., 2009; Figures 5A and 5B; Figure S5). Consistent with the results in the murine system, reprogramming was completely abrogated when one of the GTLM factors was removed, demonstrating that all four are necessary to also induce the erythroid fate in human fibroblasts (Figures 5A and 5B). Bulk human iEPs displayed a round morphology, deep blue cytoplasm, and coarse chromatin, features of early erythroid precursors (Figure 5C). Last, gene expression analysis by qPCR revealed that GTLM factors robustly induced the co-ordinated expression of several erythroid cell-specific genes, including embryonic (HBE1), fetal (HBG1/2), and adult hemoglobins (HBB and HBA1/2) (Figure 5D). As observed in the murine setting, human iEPs expressed 7.5 times more fetal HBG1/2 and 20.7 times more HBE1 than adult HBB, supporting the previous finding that GTLM induces an embryonic erythroid program. Fibroblast-specific genes were less effectively downregulated upon reprogramming than in mouse TTFs, which is likely reflective of the cellular heterogeneity of the bulk populations analyzed.

Figure 5
Induction of Erythroid Progenitors from Human Foreskin Fibroblasts by Gata1, Tal1, Lmo2, and c-Myc

Altogether, we provide proof-of-principle evidence that iEPs can be generated from human fibroblasts by overexpression of GTLM, reproducing the findings in the murine setting and underscoring a conserved transcriptional program instructing the erythroid cell fate in mammalian cells.


The identification of the minimal set of factors required to instruct erythroid lineage fate could provide a strategy to study and recapitulate erythropoiesis in vitro for medical purposes. We show that murine and human fibroblasts can be rapidly and directly converted into erythroid progenitor/precursor cells by forced expression of Gata1, Tal1, Lmo2, and c-Myc. iEPs exhibit properties of bona fide erythroid cells, such as morphology, gene expression, and colony-forming capacity, suggesting that GTLM constitute the core network of the erythroid program capable of orchestrating the battery of factors necessary for normal RBC development.

In erythroid cells, GATA1 and TAL1 are known to assemble within multimeric protein complexes, also including the adaptor molecules LMO2 and LDB1 (Osada et al., 1997) and reviewed by Love et al. (2014). These complexes show widespread binding at erythroid genes and erythroid enhancer elements, functioning as primary mediators of global erythroid gene activation (Li et al., 2013). In addition, the ability of LDB1 protein complexes to oligomerize facilitates long-range associations between promoters and enhancers, which is essential for β-globin, Myb, and Epb4.2 gene expression (Song et al., 2007). Ldb1 is already expressed in fibroblasts (Figure S3D), which likely explains why it is not additionally necessary for iEP generation. Thus, our results suggest that the key event leading to induction of the erythroid fate is the assembly of the GATA1/TAL1/LMO2/LDB1 complex.

The role of c-Myc in iEP generation is more ambiguous because its protein is not known to interact with GATA1, TAL1, or LMO2. c-Myc is one of the of the four transcription factors used to originally generate induced pluripotent stem cells (iPSCs) from fibroblasts (Takahashi and Yamanaka, 2006). In that context, c-Myc has been shown to enhance the early steps of reprograming by repressing fibroblast-specific genes and upregulating the metabolic program of the embryonic state (Sridharan et al., 2009). For iPSC generation, c-Myc can be omitted if p53-null fibroblasts are used (Hong et al., 2009). In our system, reprogramming of p53-null fibroblasts greatly enhanced efficiency but did not allow reprogramming without c-Myc (data not shown), which suggests that c-Myc has a role beyond that shown in iPSC generation. Another indication of c-Myc’s erythroid-specific requirement is highlighted by the fact that epiblast-restricted c-Myc-null mouse embryos die at E12 from severe anemia (Dubois et al., 2008). Thus, c-Myc is likely to have more than just a cell proliferation function during erythroid reprogramming, and future investigation will be needed to determine its precise role in iEP generation. In addition, the observation that iEP reprogramming is enhanced by increasing the amount of Gata1 while increasing c-Myc, instead, blocks development of hemoglobinized cells (Figure 2H) demonstrates that GTLM factor stoichiometry is critical and can be further optimized.

Since the advent of iPSC technology in 2006, many laboratories have been screening for factor combinations that can instruct cell fate changes with the main purpose of generating relevant cell types for regenerative medicine. Although non-hematopoietic cells have never been directly reprogrammed to erythroid cells before, different combinations of transcription factors have been shown to reprogram somatic cells to hematopoietic cells, including hematopoietic stem cells and downstream precursor cells (Doulatov et al., 2013, Kulessa et al., 1995, Pereira et al., 2013, Pulecio et al., 2014, Riddell et al., 2014). Recently, Batta et al. (2014) described the generation of hematopoietic progenitors with subsequent erythroid potential by ectopic expression of Erg, Gata2, Lmo2, Runx1c, and Scl in murine fibroblasts. However, for the factor combinations to be relevant for determining the core transcriptional machinery directing RBC development, it is necessary to reprogram directly to erythroid-restricted progenitor cells. In this regard, the most informative study is a report demonstrating that differentiated murine B cells can be reprogrammed to erythroid-like cells by forced expression of Gata1, Scl, and C/EBPα (Sadahira et al., 2012). Gata1 and Scl were sufficient for reprogramming, but C/EBPα enhanced the process by inactivating Pax6, a critical transcription factor for B cell differentiation. A second study, focusing on molecular mechanisms controlling the progression from hematopoiesis to erythropoiesis during embryogenesis, reported that a combination of five transcription factors, Scl, Lmo2, Gata2, Ldb1, and E2A, together with the inhibition of the fibroblast growth factor (FGF) pathway, directly induced erythroid differentiation in the pluripotent chicken epiblast (Weng and Sheng, 2014). Only Scl + Lmo2 and FGF inhibition were sufficient to convert nascent mesoderm cells in later stages of the chicken embryo, suggesting distinct factor requirements depending on the plasticity of the starting population. Gata1 and Klf1 could replace Gata2 in the transcription factor cocktail, but the efficiency was reduced. This study, however, investigated direct erythroid induction from pluripotent cell populations and not terminally differentiated cells, thus illustrating a case of lineage specification rather than transdifferentiation. Taken together, despite the diverse nature of starting cell sources, most of these studies employ the Gata factor family, Scl/Tal1, and/or Lmo2 as conversion factors, validating the essential role of this complex in establishing erythroid lineage identity.

Another common feature of the aforementioned reports and our findings is the detection of embryonic and fetal globins. The inability to express adult globin has also been observed in erythrocytes generated in vitro from pluripotent stem cells or cord blood progenitors, which poses a major barrier to their clinical use (reviewed by Anstee et al. (2012). Together, these observations support the hypothesis that transcription factor-driven reprogramming to blood follows early developmental steps and suggests that the addition of factors mediating the switch to adult globin may be necessary for maturation. Consistent with this, we found that overexpression of Klf1 or Myb changed the globin expression pattern in single iEPs from predominantly embryonic to mainly adult. Although both Klf1 and Myb were found to be dispensable during the initial screening for erythroid fate conversion factors, they were later found to induce adult globin expression. Myb is uniquely expressed during definitive erythropoiesis, and its dysregulation is associated with the persistence of embryonic and fetal hemoglobins in Trisomy 13 (Sankaran et al., 2011, Tober et al., 2008). Klf1, in contrast, is required for both primitive and definitive erythropoiesis and plays a crucial role in regulating the expression of adult and embryonic globins (Hodge et al., 2006).

Here we show direct and robust red cell fate conversion from a mammalian, non-hematopoietic, differentiated somatic cell. We demonstrate that transcription factor-mediated direct conversion can be employed as an unambiguous method to define the core transcriptional program of a cell type, allowing the distinction of “fate conversion factors” versus “maturation factors.” This method can be applied to define the factors required for human adult erythropoiesis and model disease. Furthermore, the GTLM factors could potentially be used to enhance methods for in vitro production of erythrocytes for personalized transfusion medicine.

Experimental Procedures


Epor-Cre R26-eYFP mice were on a C57Bl/6 background and have been described previously (Heinrich et al., 2004, Singbrant et al., 2011). WT C57BL/6 mice were purchased from Taconic. All animal experiments were carried out in accordance with Lund University’s ethical regulations (Ethical Permit M253-12).

Establishment of Fibroblast Cultures from Mouse Tail Tips

The procedure utilized was a modified version of the protocol by Takahashi and Yamanaka (2006). Tail tips were taken from 6- to 8-week-old mice. The fur was dissected out, and the remaining tissue was manually minced with a sterile scalpel into 1-cm-long pieces and further dissected in the presence of 0.4% trypsin (Thermo Scientific). Tissue pieces were plated on cell culture dishes pre-coated with 0.1% gelatin and incubated in fibroblast medium for 5 days at 37°C, 5% CO2, 4% O2. Confluent cultures were passaged twice before hematopoietic lineage depletion and then used for iEP generation. TTF cultures were derived in DMEM supplemented with 15% fetal calf serum (FCS) (Thermo Scientific), 2 mM L-glutamine (Gibco), 1% nonessential amino acids (Gibco), and 1% antibiotics (fibroblast expansion medium [FEX]).

Molecular Cloning and Production of Retrovirus

For the initial screening, coding regions for candidate genes (Table S1) were amplified and cloned into the pMXs retroviral vector backbone using BamHI and NotI restriction sites. The details of the primer sequences are provided in Table S4. For hemoglobin switching experiments, coding regions of Klf1 and Myb were digested out of the pMXs vector using BamHI and SalI and cloned into the pWZL-blast retroviral vector backbone using the same restriction sites. For retrovirus production, 2 × 106 Phoenix GP cells were seeded per 100-mm dish without antibodies. The next day, the medium was changed to pure DMEM, and pMXs plasmids and the Ecopac vector were transfected using FuGENE 6 transfection reagent (Promega). 27 μl of FuGENE 6 was added to 400 μl of DMEM and incubated for 5 min at room temperature. 6 μg of pMXs vector and 3 μg of Ecopac vector were diluted in 25 μl of DMEM and added to the previous mixture, which was incubated for 20 min at room temperature. The DNA/FuGENE 6 mixture was added drop by drop onto the Phoenix GP cells, which were incubated overnight at 37°C, 5% CO2. The next day, the medium was changed to DMEM + 20% FCS. Viral supernatants were harvested 48 hr after transfection, filtered through a 0.45-μm filter, and frozen for later use.

iEP Generation

Mouse iEP Generation

TTFs were seeded at 1 × 104 cells/cm2 on 0.1% gelatin pre-coated dishes and infected 24 hr thereafter with non-concentrated, virus-containing supernatants supplemented with 4 μg/ml of Polybrene (Merck Millipore). Virus supernatant and FEX were combined at a ratio of 0.6:1 during transduction. After 4 hr, the medium was changed to serum-free expansion medium (SFEM) (STEMCELL Technologies) supplemented with 1% antibiotics, 100 ng/ml mSCF, 10 ng/ml mIL-3, 2 U/ml recombinant human Erythropoietin (rhEPO), and 100 nM dexamethasone and cultured at 37°C, 5% CO2 and hypoxia (4% O2). Emerging murine iEP colonies were scored after 5–8 days after transduction.

Human iEP Generation

For human iEPs, 1 × 105 HFFs were infected three times with several combinations of murine retroviruses at 1,800 rpm for 45 min at 32°C in the presence of 4 μg/ml Polybrene and cultured at 32°C for 12 hr between infections. The culture medium containing the viruses was changed the day thereafter for reprogramming medium (Iscove’s modified Delbecco’s medium [IMDM], 0.4% albumin, 1% human serum supplemented with 1% antibiotics, insulin-transferring-selenium [ITS] (Gibco), 100 ng/ml hSCF, 10 ng/ml hIL-3, 2 U/ml rhEPO, and 100 nM dexamethasone) and cultured at 37°C, 5%CO2, and 4% O2.

Establishment of Mouse Fibroblast Cultures with Constitutive Expression of Klf1 or Myb

TTFs were seeded and transduced with pWZL-blast-Klf1 or pWZL-blast-Myb as described under iEP Generation. 48 hr after transduction, cells were passaged and cultured in FEX supplemented with 4 μg/ml blasticidin antibiotic (InvivoGen). The whole medium was replaced with fresh antibiotic-containing medium every 3 days. When a population of resistant TTFs was obtained, cells were seeded accordingly and transduced with pMX vectors to generate iEPs. Blasticidin was added to the medium during reprogramming.

Flow Cytometry

Clusters of round cells were collected through gentle pipetting, and untransduced fibroblasts were trypsinized. Cells were resuspended in 100 μl of PBS + 2% FCS and stained for 20 min at 4°C in the dark. The list of antibodies and dilutions can be found in the Supplemental Experimental Procedures. After staining, cells were washed and resuspended in 300 μl of PBS with 2% FCS. To assess cell viability, cells were also stained with 1:100 DAPI. Compensation controls were set with murine bone marrow cells. The analysis of mouse cells was performed using FACSCanto II (Becton Dickinson), and data were analyzed using FlowJo v10 software. The analysis of human cells was performed using a Gallios flow cytometer (Becton Dickinson).

RNA Isolation and qPCR

Total RNA was isolated with the RNAeasy kit (QIAGEN) according to the manufacturer’s guidelines. RNA was subjected to cDNA synthesis using Superscript III (Invitrogen). Quantitative PCR analysis was performed in triplicates for each sample and Taqman gene expression assays (Applied Biosystems) or PrimeTime qPCR assays (Integrated DNA Technologies). For each reaction, 10 μl of Taqman master mix, 4 μl of sterile water, and 1 μl of the primer assay were mixed. Then, 5 μl of cDNA (<40 ng/μl) was added to each well. The signal was detected with a 7900HT RT-PCR instrument (Applied Biosystems). The primers used are listed in Tables S2 and S3.

Colony Assay

For colony-forming assays, MethoCult M3236 (STEMCELL Technologies) was used, which was supplemented with 5 U/ml rhEPO, 50 ng/ml mSCF, 100 nM of dexamethasone, and 1% antibiotics. Cells were resuspended in FCS and added to the MethoCult mixture. All experiments were performed in triplicate. Colonies were scored on day 8.

Global Gene Expression Analysis

Total RNA was isolated from red iEP, non-red iEP, FL colony, BM colony, and fibroblasts and hybridized to a Mouse Gene 2.1 ST array (Affymetrix) according to the manufacturer’s protocol. The arrays were robust multi-array average (RMA)-normalized, and differentially expressed genes were identified using LIMMA (25605792). The microarray data have been deposited in NCBI’s GEO (Edgar et al., 2002) and are accessible through GEO: GSE73344.

Single-Cell qRT-qPCR and Data Analysis

Single YFP+ Ter119+ cells were sorted into 96-well PCR plates (Sarstedt) containing 4 μl of lysis buffer. Next, target-specific pre-amplification was performed using the CellsDirect one-step qRT-PCR kit (Life Technologies). Pre-amplified samples were diluted 1:5 and run on Fluidigm 96.96 arrays on a Biomark device (Fluidigm) together with the Taqman assays listed in Table S2 and reagents according to the manufacturer’s instructions. The data were analyzed using the Single Cell Expression Visualizer web tool ( (Lang et al., 2015). See the Supplemental Experimental Procedures for details.

Statistical Evaluation

Statistical analyses of all endpoints were performed using unpaired Student’s t test (Figure 2F) and two-way ANOVA test (Dunnett correction for multiple comparisons, 95% confidence intervals) (Figures 1C, C,2J,2J, J,4A4A and 4B, and and5B)5B) using GraphPad Prism 6.

Author Contributions

Conceptualization, S.C.G., K.D., K.S., S. Singbrant, V.G.S., and J.F.; Methodology, S.C.G., K.D., K.S., V.G.S., and J.F.; Software and Formal Analysis, S. Soneji; Investigation – Murine Data, S.C.G., K.D., and K.S.; Investigation – Human Data, J.P. and S.C.G.; Library Generation, V.R.E. and J.F.; Single-Cell qRT-PCR Conceptualization and Analysis, M.S., S.C.G., and G.K., Writing – Original Draft, S.C.G. and J.F.; Writing – Review & Editing, S.C.G., V.G.S., S. Singbrant., K.S., S. Soneji, V.R.E., J.P., A.R., C.W., and J.F.; Funding Acquisition, A.R. and J.F.; Resources, C.W., A.R., and J.F.; Supervision, J.F.


We thank Evelyn Wang and Gregory Hyde (Whitehead Institute, Cambridge, MA) for cloning and Harvey Lodish (Whitehead Institute) for providing many of the plasmids used for generating the retroviral library. We thank Jacob Ulirsch (Broad Institute of MIT and Harvard) for early contributions to the global gene expression analysis. We thank Roger Rönn (Department of Molecular Medicine and Gene Therapy) for input on the manuscript and Zhi Ma (Lund Stem Cell Center) for flow cytometry assistance. This work was supported by the Ragnar Söderberg Foundation (to J.F.); the Swedish Research Council (to J.F.); Stiftelsen Olle Engkvist Byggmästare (to J.F.); the Swedish Foundation for Strategic Research (to J.F.); Åke Wiberg’s Foundation (to J.F.); a Marie Curie integration grant (to J.F.); and grants from MINECO (SAF2012-33526) (to A.R.), ISCIII – FEDER (RETICS RD12/0019/0019, RD12/0019/0034) (to A.R.), Fundació Marató de TV3 (121430) (to A.R.), and AGAUR (2014-SGR-1460, 2014-SGR-1570) (to A.R.). J.P. was partially supported by a Juan de la Cierva postdoctoral fellowship (JCI-2012-15293).


Published: June 2, 2016


Supplemental Information includes Supplemental Experimental Procedures, five figures, and four tables and can be found with this article online at

Accession Numbers

The microarray data have been deposited in NCBI’s Gene Expression Omnibus (Edgar et al., 2002) and are accessible through GEO: GSE73344.

Supplemental Information

Document S1. Supplemental Experimental Procedures, Figures S1–S5, and Tables S1–S4:
Document S2. Article plus Supplemental Information:


Anstee D.J., Gampel A., Toye A.M. Ex-vivo generation of human red cells for transfusion. Curr. Opin. Hematol. 2012;19:163–169. [PubMed]
Batta K., Florkowska M., Kouskoff V., Lacaud G. Direct reprogramming of murine fibroblasts to hematopoietic progenitor cells. Cell Rep. 2014;9:1871–1884. [PubMed]
Cantor A.B., Orkin S.H. Transcriptional regulation of erythropoiesis: an affair involving multiple partners. Oncogene. 2002;21:3368–3376. [PubMed]
Chen K., Liu J., Heck S., Chasis J.A., An X., Mohandas N. Resolving the distinct stages in erythroid differentiation based on dynamic changes in membrane protein expression during erythropoiesis. Proc. Natl. Acad. Sci. USA. 2009;106:17413–17418. [PubMed]
Doulatov S., Vo L.T., Chou S.S., Kim P.G., Arora N., Li H., Hadland B.K., Bernstein I.D., Collins J.J., Zon L.I., Daley G.Q. Induction of multipotential hematopoietic progenitors from human pluripotent stem cells via respecification of lineage-restricted precursors. Cell Stem Cell. 2013;13:459–470. [PubMed]
Dubois N.C., Adolphe C., Ehninger A., Wang R.A., Robertson E.J., Trumpp A. Placental rescue reveals a sole requirement for c-Myc in embryonic erythroblast survival and hematopoietic stem cell function. Development. 2008;135:2455–2465. [PubMed]
Edgar R., Domrachev M., Lash A.E. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 2002;30:207–210. [PubMed]
Heinrich A.C., Pelanda R., Klingmüller U. A mouse model for visualization and conditional mutations in the erythroid lineage. Blood. 2004;104:659–666. [PubMed]
Hodge D., Coghill E., Keys J., Maguire T., Hartmann B., McDowall A., Weiss M., Grimmond S., Perkins A. A global role for EKLF in definitive and primitive erythropoiesis. Blood. 2006;107:3359–3370. [PubMed]
Hong H., Takahashi K., Ichisaka T., Aoi T., Kanagawa O., Nakagawa M., Okita K., Yamanaka S. Suppression of induced pluripotent stem cell generation by the p53-p21 pathway. Nature. 2009;460:1132–1135. [PubMed]
Huang W., Sherman B.T., Lempicki R.A. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 2009;37:1–13. [PubMed]
Huang W., Sherman B.T., Lempicki R.A. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc. 2009;4:44–57. [PubMed]
Huang P., He Z., Ji S., Sun H., Xiang D., Liu C., Hu Y., Wang X., Hui L. Induction of functional hepatocyte-like cells from mouse fibroblasts by defined factors. Nature. 2011;475:386–389. [PubMed]
Ieda M., Fu J.D., Delgado-Olguin P., Vedantham V., Hayashi Y., Bruneau B.G., Srivastava D. Direct reprogramming of fibroblasts into functional cardiomyocytes by defined factors. Cell. 2010;142:375–386. [PubMed]
Jopling C., Boue S., Izpisua Belmonte J.C. Dedifferentiation, transdifferentiation and reprogramming: three routes to regeneration. Nat. Rev. Mol. Cell Biol. 2011;12:79–89. [PubMed]
Kingsley P.D., Malik J., Emerson R.L., Bushnell T.P., McGrath K.E., Bloedorn L.A., Bulger M., Palis J. “Maturational” globin switching in primary primitive erythroid cells. Blood. 2006;107:1665–1672. [PubMed]
Kingsley P.D., Greenfest-Allen E., Frame J.M., Bushnell T.P., Malik J., McGrath K.E., Stoeckert C.J., Palis J. Ontogeny of erythroid gene expression. Blood. 2013;121:e5–e13. [PubMed]
Kulessa H., Frampton J., Graf T. GATA-1 reprograms avian myelomonocytic cell lines into eosinophils, thromboblasts, and erythroblasts. Genes Dev. 1995;9:1250–1262. [PubMed]
Lang S., Ugale A., Erlandsson E., Karlsson G., Bryder D., Soneji S. SCExV: a webtool for the analysis and visualisation of single cell qRT-PCR data. BMC Bioinformatics. 2015;16:320. [PubMed]
Li L., Freudenberg J., Cui K., Dale R., Song S.H., Dean A., Zhao K., Jothi R., Love P.E. Ldb1-nucleated transcription complexes function as primary mediators of global erythroid gene activation. Blood. 2013;121:4575–4585. [PubMed]
Love P.E., Warzecha C., Li L. Ldb1 complexes: the new master regulators of erythroid gene transcription. Trends Genet. 2014;30:1–9. [PubMed]
McGrath K.E., Frame J.M., Fromm G.J., Koniski A.D., Kingsley P.D., Little J., Bulger M., Palis J. A transient definitive erythroid lineage with unique regulation of the β-globin locus in the mammalian embryo. Blood. 2011;117:4600–4608. [PubMed]
Mead P.E., Deconinck A.E., Huber T.L., Orkin S.H., Zon L.I. Primitive erythropoiesis in the Xenopus embryo: the synergistic role of LMO-2, SCL and GATA-binding proteins. Development. 2001;128:2301–2308. [PubMed]
Osada H., Grutz G.G., Axelson H., Forster A., Rabbitts T.H. LIM-only protein Lmo2 forms a protein complex with erythroid transcription factor GATA-1. Leukemia. 1997;11(Suppl 3):307–312. [PubMed]
Palis J. Primitive and definitive erythropoiesis in mammals. Front. Physiol. 2014;5:3. [PubMed]
Pereira C.F., Chang B., Qiu J., Niu X., Papatsenko D., Hendry C.E., Clark N.R., Nomura-Kitabayashi A., Kovacic J.C., Ma’ayan A. Induction of a hemogenic program in mouse fibroblasts. Cell Stem Cell. 2013;13:205–218. [PubMed]
Pulecio J., Nivet E., Sancho-Martinez I., Vitaloni M., Guenechea G., Xia Y., Kurian L., Dubova I., Bueren J., Laricchia-Robbio L., Izpisua Belmonte J.C. Conversion of human fibroblasts into monocyte-like progenitor cells. Stem Cells. 2014;32:2923–2938. [PubMed]
Riddell J., Gazit R., Garrison B.S., Guo G., Saadatpour A., Mandal P.K., Ebina W., Volchkov P., Yuan G.C., Orkin S.H., Rossi D.J. Reprogramming committed murine blood cells to induced hematopoietic stem cells with defined factors. Cell. 2014;157:549–564. [PubMed]
Sadahira K., Fukuchi Y., Kunimono H., Sakurai M., Ikeda Y., Okamoto S., Nakajima H. Direct reprogramming of terminally differentiated B cells into erythroid lineage. FEBS Lett. 2012;586:3645–3652. [PubMed]
Sankaran V.G., Xu J., Ragoczy T., Ippolito G.C., Walkley C.R., Maika S.D., Fujiwara Y., Ito M., Groudine M., Bender M.A. Developmental and species-divergent globin switching are driven by BCL11A. Nature. 2009;460:1093–1097. [PubMed]
Sankaran V.G., Menne T.F., Šćepanović D., Vergilio J.A., Ji P., Kim J., Thiru P., Orkin S.H., Lander E.S., Lodish H.F. MicroRNA-15a and -16-1 act via MYB to elevate fetal hemoglobin expression in human trisomy 13. Proc. Natl. Acad. Sci. USA. 2011;108:1519–1524. [PubMed]
Sekiya S., Suzuki A. Direct conversion of mouse fibroblasts to hepatocyte-like cells by defined factors. Nature. 2011;475:390–393. [PubMed]
Shivdasani R.A., Orkin S.H. The transcriptional control of hematopoiesis. Blood. 1996;87:4025–4039. [PubMed]
Singbrant S., Russell M.R., Jovic T., Liddicoat B., Izon D.J., Purton L.E., Sims N.A., Martin T.J., Sankaran V.G., Walkley C.R. Erythropoietin couples erythropoiesis, B-lymphopoiesis, and bone homeostasis within the bone marrow microenvironment. Blood. 2011;117:5631–5642. [PubMed]
Song S.H., Hou C., Dean A. A positive role for NLI/Ldb1 in long-range beta-globin locus control region function. Mol. Cell. 2007;28:810–822. [PubMed]
Sridharan R., Tchieu J., Mason M.J., Yachechko R., Kuoy E., Horvath S., Zhou Q., Plath K. Role of the murine reprogramming factors in the induction of pluripotency. Cell. 2009;136:364–377. [PubMed]
Szabo E., Rampalli S., Risueño R.M., Schnerch A., Mitchell R., Fiebig-Comyn A., Levadoux-Martin M., Bhatia M. Direct conversion of human fibroblasts to multilineage blood progenitors. Nature. 2010;468:521–526. [PubMed]
Takahashi K. Cellular reprogramming--lowering gravity on Waddington’s epigenetic landscape. J. Cell Sci. 2012;125:2553–2560. [PubMed]
Takahashi K., Yamanaka S. Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell. 2006;126:663–676. [PubMed]
Tober J., McGrath K.E., Palis J. Primitive erythropoiesis and megakaryopoiesis in the yolk sac are independent of c-myb. Blood. 2008;111:2636–2639. [PubMed]
Trakarnsanga K., Wilson M.C., Lau W., Singleton B.K., Parsons S.F., Sakuntanaga P., Kurita R., Nakamura Y., Anstee D.J., Frayne J. Induction of adult levels of β-globin in human erythroid cells that intrinsically express embryonic or fetal globin by transduction with KLF1 and BCL11A-XL. Haematologica. 2014;99:1677–1685. [PubMed]
Tsiftsoglou A.S., Vizirianakis I.S., Strouboulis J. Erythropoiesis: model systems, molecular regulators, and developmental programs. IUBMB Life. 2009;61:800–830. [PubMed]
Vierbuchen T., Wernig M. Direct lineage conversions: unnatural but useful? Nat. Biotechnol. 2011;29:892–907. [PubMed]
Vierbuchen T., Ostermeier A., Pang Z.P., Kokubu Y., Südhof T.C., Wernig M. Direct conversion of fibroblasts to functional neurons by defined factors. Nature. 2010;463:1035–1041. [PubMed]
Weng W., Sheng G. Five transcription factors and FGF pathway inhibition efficiently induce erythroid differentiation in the epiblast. Stem Cell Reports. 2014;2:262–270. [PubMed]
Yi Z., Cohen-Barak O., Hagiwara N., Kingsley P.D., Fuchs D.A., Erickson D.T., Epner E.M., Palis J., Brilliant M.H. Sox6 directly silences epsilon globin expression in definitive erythropoiesis. PLoS Genet. 2006;2:e14. [PubMed]
Zhou D., Liu K., Sun C.W., Pawlik K.M., Townes T.M. KLF1 regulates BCL11A expression and gamma- to beta-globin gene switching. Nat. Genet. 2010;42:742–744. [PubMed]