Induced pluripotent stem cells (iPSCs) hold great promise for cell therapy. However, their low efficiency of lineage-specific differentiation and tumorigenesis severely hinder clinical translation. We hypothesized that reprogramming of somatic cells into lineage-specific progenitor cells might allow for large-scale expansion, avoiding the tumorigenesis inherent with iPSCs and simultaneously facilitating lineage-specific differentiation. Here we aimed at reprogramming rat hepatic WB cells, using four Yamanaka factors, into pancreatic progenitor cells (PPCs) or intermediate (IM) cells that have characteristics of PPCs. IM clones were selected based on their specific morphology and alkaline phosphatase activity and stably passaged under defined culture conditions. IM cells did not have iPSC properties, could be stably expanded in large quantity, and expressed all 14 genes that are used to define the PPC developmental stage. Directed differentiation of IM and WB cells by Pdx1-Ngn3-MafA (PNM) into pancreatic beta-like cells revealed that the IM cells are more susceptible to directed beta cell differentiation because of their open chromatin configuration, as demonstrated by expression of key pancreatic beta cell genes, secretion of insulin in response to glucose stimulation, and easy access to exogenous PNM proteins at the rat insulin 1 and Pdx1 promoters. This notion that IM cells are superior to their parental cells is further supported by the epigenetic demonstration of accessibility of Pdx1 and insulin 1 promoters. In conclusion, we have developed a strategy to derive and expand PPC cells from hepatic WB cells using conventional cell reprogramming. This proof-of-principal study may offer a novel, safe and effective way to generate autologous pancreatic beta cells for cell therapy of diabetes.
The incidence of type 1 diabetes (T1D) is increasing dramatically. Islet beta cell replacement therapy is one of the most promising approaches for treating and curing T1D. However, its large scale application is hampered by a shortage of islet beta cells for transplantation and requirement of life-long immune suppression (Atkinson and Eisenbarth, 2001). Reprogramming of somatic cells into induced pluripotent stem cells (iPSCs) is a recent breakthrough that holds great promise for providing autologous, lineage-specific differentiated cells for cell replacement therapy (Takahashi and Yamanaka, 2006). However, the low efficiency of lineage-specific differentiation (Cohen and Melton, 2011; Maehr et al., 2009; Sancho-Bru et al., 2011) along with teratoma formation from residual undifferentiated iPSCs (Ben-David and Benvenisty, 2011; Chambers and Studer, 2011; Naujok et al., 2011; Pera, 2011) could adversely impact clinical application. Given that the ultimate goal for translational medicine is to produce differentiated cell types on demand, a preferable approach might be to reprogram somatic cells directly to a progenitor-like stage that is highly expansible and lineage-specific, rather than all the way back to the stage of pluripotency.
Progress in developmental biology has identified a common pool of undifferentiated progenitor cells that populate the early pancreatic buds and give rise to both endocrine and exocrine cells during pancreas development (Bonner-Weir and Sharma, 2002; Edlund, 2001; Lynn et al., 2007; Wilson et al., 2003). These early progenitor cells, also known as pancreatic progenitor cells (PPCs), are defined by the expression of a set of transcription factors: Tcf2, HNF6, Foxa2, Hb9, Pdx1, Ptf1a, Nkx2.2, Nkx6.1, Sox4, Gata6, Gata4 and Sox9 (Lynn et al., 2007; Maestro et al., 2003). Some of these markers Tcf2, HNF6 and Foxa2 also are expressed in developing gut and liver and persist in the pancreatic ducts of the adult pancreas (Lynn et al., 2007). These factors form a complex network that controls PPC identity and supports endocrine cell differentiation within the developing organ (Lynn et al., 2007). Although adult pancreatic ducts harbor the pancreatic progenitor/stem cells (Bonner-Weir and Sharma, 2002), the isolation and expansion of pure PPCs has been elusive, in part due to the lack of unique surface markers that can be used for isolation and to imperfect culture protocols for maintaining PPC characteristics. The gene-expression programs that maintain progenitor cells and permit their proper differentiation are poorly understood (Weir et al., 2011). Thus, reprogramming somatic cells into PPCs may improve our understanding of these cells and facilitate clinical translation.
Classically, only rare somatic cells can be reprogrammed into iPSCs; most are partially reprogrammed to diverse intermediate states (Yamanaka, 2009). In early passages, the large population of partially reprogrammed cells expresses various lineage-specific markers showing preference for the lineage of the somatic cell's embryonic origin due to the preserved epigenetic memory (Kim et al., 2010; Maherali et al., 2007; Meissner et al., 2007; Mikkelsen et al., 2008; Silva et al., 2008; Sridharan et al., 2009). Therefore, we tested the hypothesis that pancreatic progenitor-like cells or intermediate (IM) cells that have characteristics of PPCs might be present among those partially reprogrammed cells and could be isolated and expanded under the right conditions. We reasoned that these cells may be more efficiently directed toward pancreatic beta cell differentiation than their parental cells.
Taking advantage of the preserved epigenetic memory during cell reprogramming (Kim et al., 2010) and the fact that liver and pancreas are derived from a common pool of embryonic precursor cells (Deutsch et al., 2001), we chose endoderm-derived rat hepatic WB cells (Grisham et al., 1993) as the cell source for reprogramming using retroviral transduction to deliver four Yamanaka reprogramming factors, Oct4, Sox2, Klf4 and cMyc. We successfully derived and propagated stable IM cells with a gene expression profile resembling PPCs. Furthermore, when IM cells were induced to pancreatic beta cell differentiation using three powerful pancreatic transcription factors (PTFs) in a single retroviral vector, IM cells differentiated more efficiently than the parental WB cells into functional pancreatic beta-like cells because IM cells have a more open chromatin configuration in Pdx1 and insulin 1 promoters and are more accessible to PTF proteins. Our results demonstrate the feasibility of deriving highly expansible, PPC-like IM cells via reprogramming of lineage-related cells. These studies may open a new avenue for obtaining an unlimited supply of insulin-producing cells from autologous PPC-like IM cells for cell replacement therapy of diabetes.
Generation and maintenance of IM cells
To reprogram hepatic WB cells into PPC-like IM cells, we first transduced hepatic WB cells with a combination of retroviruses encoding the gene products of murine Oct4, Sox2, Klf4 and cMyc. The timeline and selection strategy for IM clones are outlined in Fig. 1A. At day 5, the cells were re-seeded on mouse embryonic fibroblasts (MEFs) in M1 medium. Around 3 weeks, colonies with various shapes emerged that could be divided into four main types (A, B, C and D) on the basis of colony morphology (Fig. 1B, P0). Representative colonies of each type were hand-picked, expanded, and passaged in M2 medium. Fig. 1B illustrates the morphology of each colony type at two different passages (P3 and P5). Representative clones were derived from each colony type. Clones from type A, B or C colonies had no clear boundaries and were likely to be partially reprogrammed cells; clones from type D colonies had a clear boundary similar to ones giving rise to iPSCs. When these clones were stained for alkaline phosphatase (AP) at passage 6, type A, C and D clones were partially AP-positive and type B clones were negative. Because our goal was to obtain PPCs, we excluded AP-positive clones (clone A, clone C and iPSC-like clone D) and expanded and characterized the B clones. Morphologically, clone-B cells were smaller than WB cells had a high nuclear/cytoplasmic ratio, grew in monolayers and were loosely associated with each other (Fig. 1B, clone B). These morphologic features suggest that they are at an intermediate state (IM) between WB cells and iPSCs. We designated the clone-B cells as IM cells. IM cells, along with the other three clones (A, C and D) were further tested for expression of pancreatic progenitor genes (below, Fig. 2A). IM cell generation was repeated independently three times with similar results.
IM cells have now been expanded up to 35 passages without showing signs of senescence, differentiation or transformation on MEFs with M2 medium. To further explore the optimal culture conditions for long-term maintenance of IM cells, we plated cells on gelatin-coated or MEF-coated 6-well plates in either M1 or M2 culture medium. IM cells quickly differentiated at day 4 post-passage as evidenced by morphologic changes (enlarged cell size and flattened appearance) on MEF-coated wells in M1 medium or on gelatin-coated wells in M2 medium (Fig. 1C), indicating that they have to be maintained on a MEF layer with M2 medium (Fig. 1C) containing cytokines and inhibitors that prevent cell differentiation.
Characterization of IM cells
The morphological features suggest that IM cells are at various stages of de-differentiation from transitional intermediates to iPSCs. Since Yamanaka-factor-mediated reprogramming tends to rewind or dedifferentiate the hepatic cell developmental program, and liver and pancreas are developed from a common embryonic precursor cell, we tried to determine the equivalent developmental stages of the reprogrammed IM cells by examining the expression of 14 genes that are used to define cells at the PPC stage (Lynn et al., 2007) at passage 7. RT-PCR analysis revealed that, among these clones (A, B, C and D), the IM clone-B cell expression profile most closely resembled the pancreatic progenitor stage. IM cells expressed all of the 14 PPC genes (Sox9, Nkx2.2, Nkx6.1, Gata4, Gata6, Sox4, HB9, Pdx1, Ptf1a, Tcf2, HNF6, Krt19, Foxa2 and Sox17), but not Ngn3, a defining marker of pancreatic endocrine precursor cells (Fig. 2A). Thus, IM cells are at the PPC stage and have not yet become pancreatic endocrine precursors. In contrast, WB cells, though sharing the expression of several common endodermal genes (Gata4, Gata6, Sox4, and HNF6), exhibited either weak or no expression of the other pancreatic PPC genes, especially the key genes, Pdx1 and Krt19. Furthermore, the PPC gene expression profiles of IM cells at different passages (P7, P23 and P28) exhibited similar pancreatic progenitor gene expression profiles, indicating that they can be stably maintained at the PPC stage.
Because IM cells were derived in the same manner as iPSCs, we next examined pluripotency properties using traditional methods. Unlike iPSCs, IM cells stained negatively for AP and failed to form teratomas in vivo following injection of IM cells into NOD/Scid mice for 3 months (data not shown). Flow cytometric analysis for pluripotency-associated protein expression revealed that in contrast to the parental WB cells, IM cells weakly expressed Oct4, strongly expressed Sox2 and were negative for SSEA1, a key surface marker of iPSCs (Fig. 2B). RT-PCR showed that IM cells indeed expressed many pluripotency markers (Oct4, Sox2, Nanog, Klf4, cMyc and Dppa5), and the early endoderm marker, alpha-fetal protein (Afp) (Fig. 2C). In contrast, WB cells only expressed Sox2, Klf4 and cMyc, whereas the expression of Oct4, Nanog, Dppa5 and Afp was undetectable. Bisulfite sequencing analysis showed that the Oct4 promoter became more demethylated (83%) in IM cells versus WB cells (98%; Fig. 2D), but it was still highly methylated when compared to rat iPSCs (5%) (Li et al., 2009). Thus, IM cells do not have iPSC properties, as evidenced by lack of expression of pluripotency markers and AP, a highly-methylated Oct4 promoter, and failure of in vivo teratoma formation. Finally, the PPC stage of IM cells was further supported by demonstration of Sox17 and Pdx1 protein expression by immunofluorescence (IF; Fig. 2E). Overall, the data suggest that using four Yamanaka factors, with morphological criteria and AP-staining selection, we have successfully reprogrammed hepatic WB cells into stable and expansible IM cells that have the gene expression profile of PPCs.
Directed differentiation of IM cells toward pancreatic beta cells
Since IM cells have similar characteristics to PPCs, we hypothesized that they might be more susceptible to PTF-directed pancreatic beta cell differentiation. To test this hypothesis, we generated a retrovirus expressing three powerful PTFs, PNM (Pdx1-Ngn3-MafA) or PNM–eGFP containing enhanced GFP as a reporter, linked by the 2A sequence and a lentiviral vector containing the rat insulin 1 promoter (RIP)–mCherry reporter for monitoring insulin 1 (Ins1) gene expression (Fig. 3A). Separating the genes by a 2A spacer allows sequential and equimolar expression of all three PTF proteins along with eGFP (de Felipe et al., 2003), which confirms PTF expression and serves as an indicator of transduction efficiency. The expression of three PTFs following cell transduction also was confirmed by IF (data not shown). At day 4 post-transduction of IM and WB cells with PNM-eGFP and RIP-mCherry, both eGFP and mCherry were more highly expressed in WB cells (∼80%) than in IM cells (∼60%), indicating that WB cells had a higher transduction efficiency and RIP activation than IM cells (Fig. 3B), likely due to larger surface area than IM cells. Since insulin reporter activity might not accurately reflect endogenous insulin gene expression due to lack of endogenous epigenetic impact, we examined endogenous expression of the Ins1 gene by RT-PCR (Fig. 3C). Unexpectedly, Ins1 gene expression was detectable only in IM cells at day 4 post-transduction, suggesting that IM cells are more susceptible to directed differentiation than WB cells. The likely explanation is that the endogenous Ins1 gene is more accessible for exogenous PTF proteins likely due to a more open chromatin structure. Consistent with that interpretation, following treatment of PNM, IM cells, but not WB cells, gradually formed three-dimensional pancreatic islet-like cell clusters. These cell clusters emerged around day 14 post-PNM-differentiation and gradually increased in size, forming globular structures around day 24 (Fig. 3D, upper right panel, insert), a common phenomenon also seen in induced pancreatic beta cell differentiation from hepatic (Yang et al., 2002), bone marrow-derived (Cao et al., 2004), and pancreatic (Ramiya et al., 2000) stem cells. Abundant insulin protein was detected within these clusters and the surrounding PNM-IM cells as shown by insulin IF (Fig. 3D, bottom panel) than that seen in PNM-WB cells. Protein expression from beta cell-specific genes (Pdx1, Islet1, Glut2 and Nkx6.1) also was examined and Pdx1, islet1 and Glut2 were detectable by IF in both WB-PNM and IM-PNM cells at day 24 (Fig. 3E), but more intense staining was seen in IM-PNM cells than in WB-PNM cells. Interestingly, NKx6.1 protein appears mainly distributed in cytoplasm of IM-PNM cells, and only a weak signal was detected in rare IM-PNM cell nuclei when compared to rat insulinoma INS-1 cells. Nkx6.1 was undetectable in WB-PNM cells (Fig. 3E). In addition, few IM-PNM cells and rare WB-PNM cells expressed somatostatin (SST) at day 24, but glucagon (GCG) was only detected in IM-PNM cells. These results suggest that IM-PNM cells were not mature beta cells (Fig. 3F). The directed differentiation of IM cells was repeated independently three times; two of the IM clones were tested each time, and similar results were obtained.
Beta-cell-related gene expression and insulin secretion
To further support our hypothesis that IM cells might be more susceptible to PNM-directed differentiation, we examined the expression of pancreatic beta cell-specific genes Pdx1, NKx6.1 and NKx2.2, pancreatic endocrine hormone genes Ins1, insulin 2 (Ins2), glucagon (Gcg) and somatostatin (Sst), and genes related to processing insulin Pcsk1 and Pcsk2 at days 0, 4 and 24 post-induction by quantitative (q)RT-PCR (Fig. 4A). Following PNM-directed pancreatic beta cell differentiation, PNM-IM cells expressed all pancreatic endocrine genes, gradually increasing the expression of Pcsk1, Ins1, Ins2 and Sst. However, the levels of Gcg peaked at day 4 and decreased at day 24 as the cells matured toward beta cells, a pattern consistent with the notion that Gcg-expressing cells serve as beta cell precursors (Collombat et al., 2009). PNM-WB cells had a similar gene expression profile, but no detectable Gcg, Nkx6.1 or Pcsk1 expression. The expression of some genes (Gcg, Pdx1, Nkx6.1, Nkx2.2 and Pcsk2) transiently peaked at day 4, likely due to an acute response to exogenous PTFs. Of note, Ins1 gene expression in PNM-IM cells was significantly higher than in PNM-WB cells at days 4 and 24. The kinetics of pancreatic gene expression supports the hypothesis that IM cells are more prone toward beta cell differentiation than WB cells.
One of the most critical beta cell phenotypes is the capacity to synthesize and process insulin (immature beta cells), and to secrete insulin in response to glucose challenge (maturing beta cells) (MacDonald et al., 2005). IM and WB cells at days 24 and 31 post-directed differentiation were challenged with 25 mM glucose for 1 hour and insulin secretion was measured by ELISA. As shown in Fig. 4B, glucose-stimulated insulin secretion was detectable at day 24 and markedly increased 7-days post-maturation (day 31; P<0.05), but only in IM-PNM cells. To exclude the effect of osmolarity, a non-metabolizable glucose analog, 2 deoxy-D-glucose (2-DG), at the same concentration as glucose (25 mM) was used as a control and failed to trigger insulin secretion (Mauda-Havakuk et al., 2011). These results suggest that IM-PNM cells had acquired beta cell function, as they were capable of glucose-stimulated insulin secretion via a process coupling glucose sensing, uptake, and metabolism with insulin synthesis, processing and release. To further confirm the molecular basis of beta cell maturation, we analyzed the expression of genes related to beta cell function by RT-PCR, including glucose transporter 2 (Glut 2), three ATP-sensitive potassium channel proteins (Sur1, Kir6.1 and Kir6.2) essential for regulating glucose-stimulated insulin secretion (Lorenz et al., 1998; Shiota et al., 2002), and synaptosomal-associated protein of 25 kDa (Snap25), a protein that regulates insulin release (Gonelle-Gispert et al., 2000). As shown in Fig. 4C, differentiated IM-PNM cells expressed all five beta cell function-related genes at days 24 and 31, whereas WB-PNM cells lacked expression of Kir6.1 and Snap25. These results provide a molecular basis for the observation that glucose-stimulated insulin secretion was observed in IM-PNM cells, but not WB-PNM cells.
Insulin 1 promoter DNA methylation and enrichment of H3K4me3 and H3K27me3
To provide epigenetic evidence that IM cells are more effective for PNM-directed differentiation, we next examined RIP DNA methylation and enrichment by active or repressive histone 3 marks. Differentially methylated regions (DMR) can serve as an epigenetic marker for monitoring cell reprogramming (Lister et al., 2011). For instance, insulin gene expression is regulated by its DNA promoter methylation (Yang et al., 2011). To investigate the role of epigenetic changes in PNM-directed differentiation, we examined DNA methylation of the RIP by pyrosequencing. First, we asked whether RIP CpG sites (−357, −346, −288, −243 and −109) could serve as a DMR for distinguishing beta cells from non-beta cells using DNA extracted from cells with active (rat pancreatic islets) versus inactive (rat liver) Ins1 gene. As expected, the five RIP CpG sites were <30–40% methylated in islets, but highly methylated (>75%) in the liver (Fig. 5A). The CpG sites within the Ins1 DMR were highly methylated (>75%) in both WB and IM cells (Fig. 5B), consistent with the inactive status of the Ins1 gene. Following 31 days of PNM-directed differentiation and maturation, statistically significant DNA demethylation occurred at all five RIP CpG sites (Fig. 5C) in IM-PNM cells, compared to those in WB-PNM cells, consistent with the increased Ins1 gene expression (Fig. 4A). These results indicated that IM cells are more susceptible to PNM-directed differentiation toward insulin-secreting beta-like cells, suggesting that the RIP regulatory region in IM cells has a more open chromatin structure, increasing accessibility for exogenous PTF proteins.
To confirm this, we examined active and suppressive histone marks in association with the RIP in IM and WB cells by chromatin immunoprecipitation (ChIP) assay using antibodies against histone 3 tri-methylated at lysine residues 4 (H3K4me3) and 27 (H3K27me3). H3K4me3 and H3K27me3 are usually associated with active and repressed genes, respectively (Azuara et al., 2006; Bernstein et al., 2006). In contrast to WB cells (inactive Ins1 gene) but similar to INS-1 cells (active Ins1 gene), the RIP was highly associated with the active histone mark (H3K4me3) and less associated with the suppressive mark (H3K27me3) in IM cells (Fig. 5D). These results strongly suggest that the RIP regulatory regions in IM cells have an open chromatin configuration accessible to PTFs.
Open and accessible chromatin structure in rat RIP and Pdx1 promoter and enhancer in IM cells
To demonstrate that IM cells have a more open chromatin configuration than WB cells, we performed a nucleosome protection assay to compare the accessibility to PTFs at the Ins1 and Pdx1 DNA regulatory regions among WB, IM and INS-1 cells. This assay, based on the accessibility of DNA to the GpC methyltransferase (M.CviPI), provides a footprint of nucleosome occupancies (Taberlay et al., 2011). The Pdx1, MafA, NeuroD and Islet1 proteins (the latter two are direct Ngn3-target genes) physically bind to the DNA regulatory regions of Ins1 (Melloul et al., 2002a) and are crucial for establishing and maintaining pancreatic beta cell phenotype and functions. We examined GpC methyltransferase-mediated methylation, which reflects the accessibility and distribution of nucleosomes on the DNA strand, in these three cell types (Fig. 6A). As expected, in WB cells RIP between nucleotides −722 to +85 was inaccessible, exhibiting extremely low methylated GpC sites (2.4%) due to nucleosomal protection. In sharp contrast, the RIP in INS-1 cells, which actively express the Ins1 gene, was much more accessible, as all GpC sites were largely methylated (83.5%) and exhibited little protection by nucleosomes. As predicted, the RIP in IM cells was partially accessible (31% methylated GpC sites) and most of the promoter region lacked nucleosomal protection. This finding supports our hypothesis that IM cells have a more open chromatin configuration, resulting in increased accessibility to exogenous PTF proteins. Detailed analysis of DNA binding sites by Pdx1, NeuroD, Islet1 and MafA (41,Melloul et al., 2002a; Olbrot et al., 2002) reveals that these promoter elements are rarely accessible to PTF proteins in WB cells but largely accessible in IM cells. Of note, analysis of CpG site methylation confirmed the pyrosequencing data (data not shown).
Similarly, we performed a nucleosomal protection assay on selected segments of the Pdx1 enhancer (−5582 to −5356) and promoter regions (−1870 to −1686 and −593 to −168) in WB, IM, and INS-1 cells (Fig. 6B). The Pdx1 gene enhancer plays a key role in initiation and regulation of its own expression (Melloul et al., 2002b; Sharma et al., 1996). As expected, the Pdx1 enhancer is completely accessible in INS-1 cells (96.3% methylated GpC sites) but inaccessible in WB cells (0% methylated GpC sites). However, the PPC-like IM cells are partially accessible (31.2% methylated GpC sites), suggesting an open chromatin structure. Predicted binding sites for the key transcription factors (TFs)-Sox, NeuroD, E47, MafA, Foxa2 and HNF1α proteins within this region are accessible in IM cells, but not in WB cells, which explains why IM cells are more susceptible to PNM-directed beta cell differentiation. Furthermore, Pdx1 promoter regions (I and II) also had more open accessible chromatin structure in IM cells than that in WB cells (Fig. 6B), allowing access to the predicted binding sites of the Pdx1 regulatory regions interacting with TFs (Gata, Sox, Pax4/6, SP1, Foxa2, Pdx1, E-box and C/EBP). Two of the binding sites of Pdx1 itself (GTGATG: ∼598∼−592, TCAT: −388∼−385) in region I are partially to fully accessible to Pdx1 protein compared to the inaccessible chromatin structure in WB cells, providing an epigenetic basis for positive autoregulation of Pdx1 expression and maintenance of its PPC-like phenotype. The finding of open, accessible chromatin structures at the insulin and Pdx1 promoters in PPC-like IM cells provides epigenetic validation of our experimental results that PPC-like IM cells, compared to WB cells, are superior candidates for directed beta cell differentiation.
We show that hepatic cells can be directly reprogrammed into PPC-like IM cells using conventional reprogramming methods. These PPC-like IM cells are stable, can be expanded, and are distinctly different from the parental hepatic cells or iPSCs. Because we used a four-factor reprogramming strategy, we confirmed that the IM cells do not exhibit typical properties of iPSCs by gene expression, iPSC surface protein expression, epigenetic changes at the Oct4 promoter, and in vivo teratoma formation. We also verified that these cells are at an intermediate ‘pancreatic progenitor’ stage of development by demonstrating the expression of 14 genes that define the pancreatic progenitor stage, and by the analysis of chromatin structure at the RIP and Pdx1 enhancer and promoter regions. These IM cells are superior to their parental cells in PNM-directed beta cell differentiation, as evidenced by Ins1 gene expression at day 4, formation of 3D islet-like clusters, and release of insulin in response to glucose challenge. To our knowledge, this is the first report of direct reprogramming of hepatic cells into PPC-like cells that can be maintained stably and expanded indefinitely under the defined culture conditions.
Transdifferentiation (direct conversion of somatic cells into lineage-specific differentiated cells) has been demonstrated by ectopic expression of single or combined lineage-specific transcription factor(s) (Ambasudhan et al., 2011; Cao et al., 2004; Huang et al., 2011; Ieda et al., 2010; Pang et al., 2011; Son et al., 2011; Tang et al., 2006). Particularly, in vivo and in vitro conversion of hepatocytes by overexpression of PTFs is the main approach for directed transdifferentiation into pancreatic beta cells (Cao et al., 2004; Ferber et al., 2000; Horb et al., 2003). However, a major factor limiting the clinical use of transdifferentiation is that terminally differentiated cells have limited capacity to be expanded (Ieda et al., 2010; Son et al., 2011). In contrast, iPSCs are pluripotent and have an unlimited proliferative capacity (Papp and Plath, 2011; Plath and Lowry, 2011). However, the advantages of iPSCs could prove disadvantageous during induced lineage-specific cell differentiation because of the low efficiency of developing into lineage-specific cells and the unwanted consequence of teratoma formation by residual undifferentiated iPSCs. Because of the above limitations, we and others have tried to directly reprogram somatic or adult stem cells into non-pluripotent, lineage-specific progenitor cells using the approaches similar to those used to generate iPSCs. Recently, two groups reported successful derivation of neural or blood progenitors from mouse (Kim et al., 2011) or human (Szabo et al., 2010) fibroblasts using a conventional four-factor reprogramming approach, suggesting that various lineage-specific intermediate cells are indeed present among the vast number of partially reprogrammed cells. Our successful generation of PPC-like IM cells from hepatic cells further supports the notion that various lineage-specific progenitors are present and is likely to be due to the following factors: (1) The close relationship between the hepatic and pancreatic stem cell lineages during embryonic development (Deutsch et al., 2001). (2) The use of morphological criteria and negative AP-staining to select against colonies with a tendency to become iPSCs. (3) The use of special culture conditions (high-glucose medium) that appear to favor the selection of PPC-like IM cells, an approach that also may have been important for selecting neural stem cells (Kim et al., 2011) and hematopoietic stem cells (Szabo et al., 2010). During early stages of iPSC reprogramming, there is preserved epigenetic memory from the parental cells (Bar-Nur et al., 2011; Kim et al., 2010; Polo et al., 2010) with an inherited, parental cell-derived DNA methylation signature (Doi et al., 2009; Ohi et al., 2011). We found that the reprogrammed IM cells shared expression of many genes with parental WB cells. Our detailed epigenetic analysis and comparison between IM and parental WB cells, particularly histone association and nucleosome occupancies, explain the seemingly paradoxical observation that at day 4 post-PNM-directed differentiation, WB cells had stronger Ins1 reporter gene expression than IM cells, but no detectable endogenous Ins1 gene expression. Our work also provides epigenetic evidence supporting the notion that reprogramming of somatic cells directly into pancreatic progenitors can yield a nearly unlimited supply of autologous stem cells for generating pancreatic beta cells.
We have successfully generated PPC-like IM cells from hepatic WB cells. IM cells can be expanded indefinitely and stably passaged without overtly changing their phenotype. More importantly, PPC-like IM cells are highly effective for the directed pancreatic beta cell differentiation. We also provide epigenetic evidence confirming that this effect is due to a more accessible chromatin structure that exposes potential binding sites for key transcription factors (TFs) in the Pdx1 DNA regulatory regions. Like INS-1 cells, the regulatory regions of Pdx1 in PPC-like IM cells were either loosely associated with nucleosomes or nucleosome free, but they were completely protected in WB cells (Fig. 6B). The open chromatin structures in PPC-like cells allow numerous potential DNA binding sites to become accessible to the key TFs, such as Sox17 (Lefebvre et al., 2007) and Foxa2 (Melloul et al., 2002b), which are important for developing endoderm, early liver and pancreas. More importantly, accessibility of the Pdx1 regulatory region to key PTFs including Pdx1 (Spence and Wells, 2007), NeuroD (Melloul et al., 2002b), Pax4, Pax6 and MafA (Ben-Othman et al., 2013) also allows the IM cells to be more efficient as pancreatic progenitors, instead of hepatic progenitors, in promoting beta cell differentiation. This conclusion is supported by the following: (1) Compared to WB cells, IM cells express all 14 markers that define the pancreatic progenitor stage, and (2) IM cells also express more pluripotent markers and less hepatic markers (data not shown) than the parental WB cells, making them less likely to serve as hepatic progenitors. The overall gene expression profile, results from PNM-directed beta cell differentiation, and epigenetic evidence at the Pdx1 and insulin promoters confirm that the reprogrammed PPC-like IM cells are more efficient as pancreatic progenitors than parental WB cells.
Despite the great advantages of utilizing PNM as an inducer for directing beta cell differentiation in vitro, the purpose here was not to generate mature beta cells, since the delivery of all three genes in one construct made it impossible for temporal changes (or shut-down) of gene expression to take place. Ngn3 is expressed transiently during pancreatic endocrine cell specification (Gradwohl et al., 2000; Jenny et al., 2002) and persistent expression of Ngn3 causes cells exit to the cell cycle, making them unable to continue cell proliferation (unpublished observation). Therefore, future studies for PNM-directed beta cell differentiation from PPC should deliver Ngn3 in a transient fashion so as to mimic the temporal program of pancreatic beta cell development (Johansson et al., 2007; Soria, 2001). Because of the inability to terminate Ngn3 expression in PNM-treated cells, these cells were unable to continue proliferating and completely maturing. Nevertheless, they served the purpose for the present studies because they transiently exhibited some key beta cell properties.
In summary, our studies provide a proof-of-concept that reprogramming of somatic cells directly into pancreatic lineage-specific progenitor cells is feasible and it could facilitate the clinical translation of stem cell therapy by offering a safe and effective strategy to generate pancreatic beta cells for cell replacement therapy of diabetes.
Materials and Methods
Cell culture, cell transduction and colony derivation
‘Insulinoma’ INS-1 cells were cultured as previously described (Cao et al., 2004). Hepatic WB cells at passage 12 (a generous gift of Dr William Coleman, University of North Carolina at Chapel Hill) were maintained in RPMI 1640 medium supplemented with 10% FCS and 11.1 mM D-glucose (Cao et al., 2004; Grisham et al., 1993). At passage 15 they were transduced by retroviruses encoding mouse Oct4, Sox2, Klf4 and cMyc (Stemgent), according to the manufacturer's instructions. Five days later, 200,000 transduced WB cells were seeded into a 10 cm dish pre-coated with inactivated CF1 mouse embryonic fibroblasts (MEFs) (GlobalStem) and incubated with M1 medium: knockout™ DMEM, 15% knockout serum replacement (KSR), 1% glutamax, 1% non-essential amino acids (NEAAs), 1% penicillin/streptomycin (P/S), 0.1 mM β-mercaptoethanol (β-ME) and 10 ng/ml mLIF (Invitrogen). After three weeks, four morphologically different types of colonies (A–D) were hand-picked for expansion on MEFs in M2 medium: DMEM/F12, 20% KSR, 1% glutamax, 1% NEAAs, 1% P/S, 0.1 mM β-ME and 10 ng/ml bFGF (Invitrogen), supplemented with 0.5 µM MEK inhibitor PD0325901 and 3 µM GSK3β inhibitor CHIR99021 (Stemgent). After five passages, cells from the expansible colonies were subjected to AP staining according to the manufacturer's protocol using the AP Detection Kit (Stemgent) for pluripotency detection.
Construction of PTF constructs and retroviral production
Retroviral vectors for Pdx1-Ngn3-MafA (PNM) and Pdx1-Ngn3-MafA with eGFP and a lentiviral vector for rat insulin 1 promoter–mCherry were constructed. First, we generated an oligonucleotide comprising the 2A sequence (synthesized by Integrated DNA Technologies), which was cloned into the pCR2.1-TOPO vector (Invitrogen) PCR product site. Next, the rat MafA gene was amplified from genomic DNA by PCR to replace its translational termination codon with a specific restriction site and the human Pdx1 gene was excised from pCMV-Sport6-Pdx1 (Open Biosystems). Then, the rMafA and hPdx1 genes were cloned upstream and downstream of the 2A sequence of pCR2.1-TOPO, respectively, to generate pCR2.1-TOPO-rMafA-2A-hPdx1. Next, Ngn3 was amplified from mouse genomic DNA by PCR to replace its translational termination codon with a specific restriction site, and also cloned upstream of the 2A sequence of the above mentioned pCR2.1-TOPO vector. Then, the mNgn3-2A sequence was excised and cloned upstream of the MafA gene of pCR2.1-TOPO-rMafA-2A-hPdx1. Finally, the mNgn3-2A-rMafA-2A-hPdx1 sequence was excised from the pCR2.1-TOPO vector and cloned into the pMXs retroviral vector to generate a construct designated as PNM. To monitor the transgene expression easily, we also introduced the eGFP gene downstream of the Pdx1 sequence to generate pMXs-mNgn3-2A-rMafA-2A-hPdx1-2A-eGFP using the aforementioned method (a diagram in Fig. 3A). The day before retroviral packaging, 2×106 293 FT cells were plated on 100 mm dishes, and incubated overnight at 37°C, in a 5% CO2 atmosphere. At day 0, the cells were transfected with 5 µg pMXs-PNM or PNM-eGFP, 3.3 µg pCL-GagPol, and 1.7 µg pHCMV-VSV-G using TransIT®-293 Transfection Reagent (Mirus). At day 1, the culture medium was replaced with fresh 10% FCS DMEM (Invitrogen). At day 3, the supernatants were collected and filtered through a 0.45 µm filter. The supernatants were centrifuged overnight at 4°C, 8000 g. Supernatants were discarded and viruses resuspended with 4 ml fresh 10% FCS DMEM.
RIP-mCherry construct and lentiviral production
The RIP-driven mCherry lentiviral vector (pTYF-RIP-mCherry) was constructed in several steps. First, the coding sequence for mCherry was amplified by PCR from the mCherry-LacRep vector (Addgene). Next, the mCherry coding sequence was inserted into our previously described RIP-eGFP (Cao et al., 2004) directly downstream of the RIP using the EcoRI/XhoI restriction sites. Finally, the RIP-driven mCherry DNA sequence was excised from the RIP-eGFP vector and inserted into the pTYF-linker vector using the BamHI/KpnI restriction sites. Lentiviruses were produced as previously described (Chang and Zaiss, 2002).
Cells (5×106 cells/mouse) were injected subcutaneously into the back of non-obese diabetic/severe combined immunocompromised mice (NOD/Scid, n = 6/group). Around 12 weeks, mice were euthanized when a visible tumor mass was detected and confirmed to be a teratoma by histological analysis. The remaining mice without visible tumors were euthanized at week 16. All mice were examined for teratoma formation by gross and microscopic inspection. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of University of Florida and followed NIH guidelines.
RT-PCR and quantitative RT-PCR
Total RNA was extracted with TRIzol reagent and digested with DNase (Invitrogen) to remove contaminating genomic DNA. RT-PCR was performed as described (Yang et al., 2002). All qRT-PCR reactions were performed in triplicate on an ABI 7900HT (Applied Biosystem, AB) with SYBR Green PCR Master Mix (AB). Expression data were normalized relative to β-actin transcript level. The fold change for each gene was calculated by the 2−ΔΔCt method. Results were confirmed using cDNA from at least three independent experiments. The qRT-PCR conditions were as follows: initial denaturation at 95°C for 5 minutes followed by 50 cycles of 15 seconds at 95°C, 30 seconds at 56°C and 30 seconds at 72°C. The primer sequences are available upon request.
WB and IM cells were analyzed for Oct4, Sox2 and SSEA1, using Human/Mouse Embryonic Stem Cell Multi-Color Flow Cytometry Kit (R&D Systems) according to the manufacturer's instructions. The corresponding isotype-matched antibodies were used as negative controls. Cells (10,000 events per sample) were acquired on a FACS Calibur flow cytometer (BD Biosciences) and data were analyzed using FCS express 4 plus software.
WB and IM cells were grown on 6-well plates and subjected to IF for expression of Pdx1 and Sox17. For the characterization of PNM-differentiated WB and IM cells, cytospin slides were prepared from WB-PNM- and IM-PNM-treated cells at day 24 and INS-1 cells were used as positive controls and subjected to IF, as previously described (Tang et al., 2004). The cells were incubated with polyclonal rabbit anti-Sox17 (1∶200; Santa Cruz, sc-20099), polyclonal rabbit anti-Pdx1 (1∶400; Abcam, ab79388), polyclonal guinea pig anti-insulin (1∶100) (Dako, A0564), anti-Islet1 (1∶100), anti-Glut2 (1∶100), anti-Nkx6.1 (1∶100), anti-GCG (1∶100), and anti-SST (1∶100; Santa Cruz) overnight at 4°C, followed by the respective secondary antibodies: goat anti-rabbit AF488, goat anti-rabbit AF594 and goat anti-guinea pig AF488 (1∶250; Invitrogen, A11008, A11037 and A11073). Nuclei were stained with DAPI (Sigma, Aldrich). Negative staining controls were carried out by replacing the primary antibody with rabbit or guinea pig sera.
Genomic DNA was isolated from the cells and modified with sodium bisulfite using Epitect Bisulfite Kit (Qiagen) according to the manufacturer's instructions. Amplified products were cloned into pGEM®-T Vector System (Promega). Randomly selected clones were sequenced with the T7 forward primer (Sigma-Genosys) for each gene and analyzed for methylation changes at CpG sites of the Oct4 promoter (−2318∼−1980 and −1762∼−1408), and GpC sites of the RIP (−722∼+85).
Directed differentiation of IM cells toward pancreatic beta cells
For directed differentiation, 3×105 IM and WB cells per well were seeded on a 6-well plate and incubated overnight at 37°C in a 5% CO2 atmosphere. Prior to plating IM cells, the plate was pre-seeded with MEFs. The cells were first co-transduced with pMXs-PNM-eGFP for confirming transduction efficiency and functionality of PNM (Fig. 3A,B). In subsequent experiments, IM and WB cells were transduced only with PNM. The next day, the medium was replaced with fresh high-glucose (25 mM) DMEM plus 10% FCS, and thereafter, the medium was changed every 3 days. At days 4 and 24, total RNA was collected for qRT-PCR analysis. After 24 days, the cells were further matured for 7 days in M3 medium: high-glucose DMEM supplemented with 10% FCS, 0.1 mM glucagon-like peptide 1, 5 mM nicotinamide and 0.01 mM exedin-4 (Sigma).
Insulin secretion assay
For static insulin secretion, differentiated cells (at days 24 and 31) were washed five times and incubated for 1 hour in Krebs-Ringer bicarbonate (KRB) buffer containing 2 mM glucose, after which supernatants were collected by centrifugation. The cells were then washed three times and incubated in KRB buffer containing 25 mM glucose or 2-DG for 1 hour. After incubation, cells were removed by centrifugation and supernatants were frozen at −80°C for later insulin assay. The insulin assay was performed with an ultrasensitive mouse insulin ELISA kit (Alpco Diagnostics) according to the manufacturer's instructions. Total cell protein was measured using the BCA Protein Assay Kit (Pierce Biotechnology) to normalize the amount of insulin secretion. Each experiment was repeated independently three times.
Methylation status of the RIP and Pdx1 enhancer and promoter were evaluated by pyrosequencing-based methylation analysis using the PSQ HS 96 Pyrosequencing System (Qiagen). All reagents were from Qiagen. Pyrosequencing was performed according to the published procedure with minor modifications (Hansmann et al., 2011).
Chromatin immunoprecipitation assay
ChIP experiments were carried out with EZ-ChIP™ chromatin immunoprecipitation kit (Millipore) according to the manufacturer's recommendations. Briefly, the cells were chemically cross-linked by adding 1/10 volume of fresh 11% formaldehyde solution. Then cells were harvested, lysed, and sonicated to shear cross-linked DNA to ∼200–1000 base pairs in length. Mouse anti-H3K4me3 (2 µg, ab8580) and rabbit anti-H3K27me3 (2 µg, ab6002, Abcam) antibodies were used for ChIP, with mouse and rabbit IgG as negative controls, respectively. Status of histone protein occupancy of the RIP was detected by DNA-PCR using primers flanking positions −281 to +22.
Nucleosome occupancy assay
Cell nucleosome occupancy was assayed as described (Taberlay et al., 2011). In brief, cell nuclei were extracted and resuspended in ice-cold nuclei extraction buffer [10 mM Tris (pH 7.4), 10 mM NaCl, 3 mM MgCl2, 0.1 mM EDTA and 0.5% NP-40, plus protease inhibitors] for 5 min on ice. Nuclei were pelleted and washed once in buffer [10 mM Tris (pH 7.4), 10 mM NaCl, 3 mM MgCl2, and 0.1 mM EDTA containing protease inhibitors]. Freshly prepared nuclei (2×105 cells) were treated with M.CviPI (NEB) according to the manufacturer's recommendation. Reactions were terminated by adding Stop Solution [20 nM Tris-HCl (pH 7.9), 600 mM NaCl, 1% SDS, 10 mM EDTA, 400 mg/ml proteinase K] and subsequently incubated at 55°C overnight. DNA was purified, bisulfite treated, and sequenced as above.
Predicted binding sites of rat Ins1 and Pdx1 promoters and the Pdx1 enhancer
Potential transcription factor binding sites in rat Ins1 and Pdx1 promoters and the Pdx1 enhancer were predicted using MatInspector in Genomatix software, which is based on weight matrices to indicate the likelihood of true binding sites (http://www.genomatix.de). By analyzing the input DNA sequences of rat Ins1 (−722 to +85), rat Pdx1 enhancer (−5690 to −5272), Pdx1 promoter (−611 to −121, region I) and (−1891 to −1671, region II) from GeneBank, only potential binding sites having weight matrices above 0.75 and a close association with endoderm, early liver, or pancreatic development were listed in Fig. 6. All binding sites of the listed PTFs in Ins1 promoter are supported by experimental evidence (Melloul et al., 2002a). The Pdx1 enhancer sequence (−5690 to −5272) contains experimentally verified Foxa2 and E47 binding sites (Melloul et al., 2002b), and this region was selected to represent the nucleosome occupancy status of the Pdx1 enhancer. A total of 58 potential transcription factor binding sites are predicted in this region. The Pdx1 promoter region I (−611 to −121) and region II sequence (−1891 to −1671) were used to represent the nucleosome occupancy status of the Pdx1 promoter. A total of 124 potential binding sites in region I and 24 binding sites in region I were predicted to be present by the software analysis.
Differences between groups were examined for statistical significance using Student's t-test. P-values of <0.05 were regarded as statistically significant.
We thank Xiaoli Wang for technical support and Dr. Donghai Wu for providing PTF plasmids.
Q.W. and H.W. participated in designing and performing experiments, analyzing data and drafting the manuscript; Y.S., S-W.L. and W.D. generated data; N.T., L-J.C., S.G.J., H.C. and W.R. provided reagents, scientific advice, discussion and participated in editing the manuscript; L.J.Y. supervised experiments, analyzed and interpreted data, drafted and edited the manuscript.
This work was supported by a Florida Bankhead-Coley Research Grant [grant number 09BW-12 to L.J.Y.]; the Lupus Research Institute to W.R. and L.J.Y.; and National Institutes of Health [grant numbers T32-AR007603 to H.W. & W.D., and R01 DK071831 to L.J.Y.]. Deposited in PMC for release after 12 months.