Generation of hepatocyte- and endocrine pancreatic-like cells from human induced endodermal progenitor cells

Multipotent Adult Progenitor Cells (MAPCs) are one potential stem cell source to generate functional hepatocytes or β-cells. However, human MAPCs have less plasticity than pluripotent stem cells (PSCs), as their ability to generate endodermal cells is not robust. Here we studied the role of 14 transcription factors (TFs) in reprogramming MAPCs to induced endodermal progenitor cells (iENDO cells), defined as cells that can be long-term expanded and differentiated to both hepatocyte- and endocrine pancreatic-like cells. We demonstrated that 14 TF-iENDO cells can be expanded for at least 20 passages, differentiate spontaneously to hepatocyte-, endocrine pancreatic-, gut tube-like cells as well as endodermal tumor formation when grafted in immunodeficient mice. Furthermore, iENDO cells can be differentiated in vitro into hepatocyte- and endocrine pancreatic-like cells. However, the pluripotency TF OCT4, which is not silenced in iENDO cells, may contribute to the incomplete differentiation to mature cells in vitro and to endodermal tumor formation in vivo. Nevertheless, the studies presented here provide evidence that reprogramming of adult stem cells to an endodermal intermediate progenitor, which can be expanded and differentiate to multiple endodermal cell types, might be a valid alternative for the use of PSCs for creation of endodermal cell types.


Introduction
In the field of regenerative medicine, the creation of functional mature hepatocytes and insulin producing pancreatic β-cells for cell based therapy of liver diseases and diabetes, respectively, are highly sought after goals. Currently, liver failure is treated by whole liver transplantation or transplantation of primary hepatocytes as a bridge to liver transplantation, while for late stage type 1 diabetes, the whole pancreas or cadaveric islet transplantation is the current treatment of choice [1][2][3]. However, scarcity of donor organs and immunorejection represent a major hurdle to treat most patients. In addition, the pharmaceutical industry is also in need for reliable drug hepatotoxicity screening models, as druginduced liver injury is one of the most prevalent causes for drug discovery failure [4]. Although primary human hepatocytes are the best source of cells for liver toxicity assessment, scarcity of donor cells, and rapid dedifferentiation of hepatocytes cultured in vitro [5][6][7][8] represent major obstacles for these studies.
An alternative to create hepatocytes or β-cells from PSCs is the direct transdifferentiation or lineage conversion of, for instance, fibroblasts into these cell lineages using combinations of transcription factors (TFs) and small molecules. Although cells with hepatocyte-like features can be generated [23][24][25], they are not fully similar to primary human hepatocytes; by contrast, glucose-stimulated insulin producing β-cells have been generated by direct transdifferentiation [26][27][28][29][30][31][32][33]. One drawback of this approach is that transdifferentiation creates post-mitotic cells that cannot be expanded and that repeated transdifferentiations from fibroblasts, which have limited expansion potential, will be required to create new populations of target cells.
Another alternative would be to generate an expandable pool of intermediate endodermal progenitor cells that can, subsequently be differentiated to mature differentiated endodermal cells. In contrast to PSCs, such endodermal progenitors might represent a safer cell source, as differentiation to non-endodermal cells would not occur. Compared to direct lineage conversion, such endodermal progenitors can be extensively expanded [25,33].
In this study we choose to use human multipotent adult progenitor cells (hMAPCs) for transdifferentiation to expandable endodermal progenitor cells (termed iENDO cells). The rationale for the use of hMAPCs as starting population was threefold: (1) hMAPCs are derived from human bone marrow and can be expanded significantly (for ±70 Population doublings (PDs)) without acquisition of genetic abnormalities[34]; (2) hMAPCs (trade name MultiStem1) are currently being used clinically in the setting of ischemic disorders and as immunomodulators without known toxicity [35][36][37] (3) although hMAPCs can differentiate in vitro and in vivo in mesodermal cell types, including endothelium, they differentiate, unlike their rodent counterparts, less robust to endodermal cell types [14,34,38,39].
We here demonstrated that using a complement of TFs, chosen based on insights from early endoderm fate specification and differentiation, hMAPCs can be reprogrammed to expandable iENDO cells, which could then be differentiated towards hepatocyte-and pancreatic β-cell-like cells in vitro and in vivo.

Transduction efficiency in hMAPC
To determine the transduction efficiency of hMAPC, we serially diluted the PLVX-IRES-eGFP lentiviral vector (0-1000μl) and transduced an equal number of hMAPC (50,000 cells/well of 12 well plate). After 3 days, we quantified eGFP expression by FACScanto flow cytometer (BD Biosciences). Data were analyzed with FlowJo Sofware (Tree Star, Ashland, OR, USA).

iENDO cell transplantation under kidney capsule of immunodeficient BALBC Rag2 -/γc -/mice
Immunodeficient BALBC Rag2 -/γc -/mice were maintained in accordance with protocols approved by the ethics committee of KU Leuven. 14 TF iENDO cells were collected from culture dishes by 0.25% trypsin, and injected under the kidney capsule of 10-12-week-old Rag2 -/γc -/mice (N = 16). After 3 weeks (N = 3) and 3 months (N = 12), mice were sacrificed, the kidneys harvested and washed with 1X PBS 2-5 times. A small part of the graft was collected in lysis buffer for RNA extraction and qRT-PCR analysis, while the remainder was fixed with 4% PFA overnight at 4˚C for immunohistochemistry and Hematoxyline and Eosin (H&E) staining. Serum from transplanted and untransplanted mice was collected and stored at -80˚C for human albumin and C-peptide analysis.

Hepatocyte organoid differentiation from hESCs
Differentiation of hESCs to definitive endoderm and hepatocyte-like cells was performed as described previously by our group [40] with addition of 0.6% DMSO between day 0 and day 12 and 2% DMSO between day 12 and day 28. On day 20, cells were washed with 1x PBS (Cat No. 10010015, Gibco, Grand Island, NY, USA), incubated with collagenase-1 (200U/ml, 300μl for 20 minutes) and detached by pipetting. Cells were collected, spun for 5 minutes at 300g and re-suspended in LDM containing 10% FBS (Cat No. F6178, Sigma-Aldrich, Saint Louis, MO, USA), Revitacell™ (1:100), (Cat No. A2644501, Gibco, Grand Island, NY, USA), 20ng/ml HGF and 2% DMSO. Cells were plated in low attachment round bottom plates at 7000 cells/well in 20μl. After 24hours, compact organoids formed, after which 100μl LDM containing HGF (20ng/ml) and 2% DMSO was added. Subsequently, medium was changed every alternate day. On day 30, medium was collected for albumin ELISA and organoids were harvested for gene expression analysis.

Pancreatic organoid differentiation from hESCs
For hESC differentiation to pancreatic β-cells, we followed the protocol described earlier by our group with minor adaptations [18]. On day 30 cells were harvested for qRT-PCR. Gene specific primers, purchased from IDT technologies, Leuven, Belgium were designed to distinguish between transgene (CDS-IRES) and endogenous (CDS-3'UTR or Exon-Exon spanning) gene expression. Gene expression graphs or heat maps were represented relative to the housekeeping gene PPIG in log scale. Heat maps were generated using Gene E software (Broad Institute of MIT and Harvard, Cambridge, MA, USA). The efficiency of primers was tested by serial dilution of cDNA and by calculating coefficient of regression (R2). An efficiency of 95-105% with an R2! 97% was considered as good (see S2-S5 Tables for list of all qRT-PCR primers used in this study).

Cell viability staining
The LIVE/DEAD1 Viability/Cytotoxicity Kit (Cat No. L3224, Molecular probes, Eugene, USA) was used according to the manufacturer recommendations. Stained organoids were observed under inverted fluorescent microscope (Axiovert, Zeiss). Live and dead cells appear as green and red, respectively.

Albumin secretion assay by ELISA
Albumin in culture supernatants and blood of grafted mice were measured using the Human Albumin ELISA Quantitation Set (Cat No.E80-129, Bethyl Laboratories, TX, USA) and the ELISA Starter Accessory Kit (Cat No. E101, Bethyl Laboratories, TX, USA) as per manufacturer's recommendations. Absorbance was measured at 450nm using Victor 3 plate reader (Model No. 1420-041, PerkinElmer, MA, USA). The results obtained were normalized with cell number.

Human C-peptide ELISA assay
Secretion of human C-peptide in serum from mice grafted with iENDO cells was analysed using the ultrasensitive human C-peptide ELISA kit (Cat No. 10-1141-01, Mercodia AB, Uppsala, Sweden) as per manufacturer's recommendations.

Immunostaining of organoids and in vivo grafts
14TF iENDO-hepatocyte organoids and iENDO-endocrine pancreatic organoids were fixed with 4% paraformaldehyde (PFA) (Cat No. P6148, Sigma-Aldrich, Saint Louis, MO, USA) overnight at 4˚C and subsequently washed with 1X PBS and stored in 70% ethanol at 4˚C. For embedding, fixed organoids were encapsulated in 2% agarose gels and stored in 70% ethanol at 4˚C until embedding. Fixed iENDO kidney grafts and 3D hepatic and pancreatic organoids were processed using the Excelsior tissue processor and paraffin embedded using the HistoStar (Thermo Scientific). Some paraffin embedded sections were stained with Hematoxyline & Eosin staining. For other sections, immunostaining was performed after melting the paraffin and rehydrating the sections, which were then washed with PBS + 0.2% Triton-X-100 (PBST) for 5 minutes. Following antigen retrieval (Dako Target Retrieval Solution (1x)) sections were permeabilised with 0.2% PBST for 5 minutes. Undifferentiated 14TF iENDO cells cultured on coverslips in 12 well plate were fixed with 4% PFA and permeabilised with 0.2% PBST), blocked with 0.2% PBST + 5% normal donkey serum, and stained overnight at 4˚C with primary antibodies and respective isotype controls in Dako diluent. Slides were washed with 1xPBST three times. Immune complexes were detected by incubation with a AlexaFluor AF488 (Green) and AF555 (Red) (1:500) coupled to secondary antibodies for 30 min at room temperature. The nuclei were visualized using Hoechst or DAPI (1:2,000). After 3 washes, slides were mounted with prolong Gold mounting medium. The signals were detected with a Nikon Eclipse Ti-S and Axioimager.Z1 microscope (Carl Zeiss). Identical exposure times were used for isotype and specific antibodies (See S6-S8 Tables, for list of primary, secondary, isotype antibodies). (Note: For cell surface protein CXCR4, stained with anti-human CXCR4 conjugated with PE antibody (1:100), cell permeabilisation and secondary antibody step was not followed).

hMAPC cell surface maker analysis by flow cytometry
hMAPC cells were detached with 0.05% trypsin-EDTA and washed with 1xPBS containing 3% FBS. After washing, between 1-2x10 5 cells were incubated for 20min in the dark at room temperature (RT) with conjugated antibodies or isotype controls (1μg/ml). A list of primary monoclonal antibodies can be found in S9 Table). 7-AAD was used as a marker to exclude dead cells from the analysis. All cells were analyzed on a FACScanto flow cytometer (BD Biosciences) with at least 10,000 recorded events. Data were analyzed with FlowJo Sofware (Tree Star, Ashland, OR, USA).

Statistical analysis
Comparisons between two groups were analysed using an unpaired 2-tailed Student's t-test. Pvalues < 0.05( Ã ), < 0.01( ÃÃ ), <0.001( ÃÃÃ ) were considered significant. Data are shown as mean and error bars represent standard error of mean (SEM) of minimum three independent experiments. All results were analysed using GraphPad prism 6 software.

14TFs can reprogram human MAPCs into induced endodermal progenitor (iENDO) cells
Initially, we transduced human MAPCs with 16 selected TFs, including the pluripotency TFs, OCT4, SOX2, KLF4, CMYC (OKSM) and TFs known to be important in mesendoderm specification (S1A Fig). Cells were maintained in endoderm differentiation medium with Activin-A/ Wnt3a/BMP4 on Matrigel coated dishes from day 4 onwards (Fig 1B). On day 4, all TFs were highly expressed except for CEBPα (S1B Fig). Between day 9 and 20, we observed morphological changes from a mesenchymal morphology to clusters of cuboidal cells in the transduced cells, but not in untransduced or PLVX-eGFP transduced cells (S1D Fig).
On day 20 cells were harvested and transcripts for endogenous mesendoderm and as well as definitive and late endodermal marker genes assessed by qRT-PCR and compared with untransduced hMAPC cultured under the same conditions and hESCs differentiated to definitive endoderm cells (day 4) and hepatocyte-/endocrine pancreatic-like cells (day 20). For genes that were part of the pool of transduced TFs, primers were used that only detect endogenous transcripts and not the transduced transcripts (denoted by an "e" in front of the name of the gene (e.g. eGATA6). We found a significant induction of several mesendoderm (S1E Fig), definitive endoderm and epithelial cell marker genes (S1F Fig). Of note, the hepatic endoderm marker genes ALB and AFP (S1G Fig) were also induced, which could be consistent with the known role of FOXA2 and HNF1α, as key regulators of hepatocyte differentiation [23,41,42]. Therefore, we repeated the studies without using the FOXA2 and HNF1α viral vectors. Human MAPCs were transduced with 14 TFs (Fig 1A) and maintained as described above (Fig 1B). Between day 9 and 20, we again observed morphological changes from a mesenchymal morphology to clusters of cuboidal cells (Fig 1C). On day 4, all TFs were expressed (S1C On day 20, reprogrammed cells were harvested and transcripts for the markers described above assessed by qRT-PCR. We again found significant induction of many mesendoderm, definitive endoderm and epithelial marker genes (Fig 1D and 1E) compared to untransduced MAPC cultured in the same medium. Levels of mesendoderm genes were signifcantly lower than in day 4 ESC endodermal progeny and levels of definitive endoderm genes approached levels of hESC-endoderm committed cells, and as hoped for, expression of more mature hepatic and pancreatic endoderm genes remained well below levels found in ESC progeny allowed to further mature to hepatocytes and pancreatic endoderm. The endogenous pluripotency genes eOCT4 and NANOG were not activated (data not shown). iENDO cells also stained positive for endodermal markers (CXCR4: 84% ± 9%, SOX17: 90% ± 16%; CK18: 85% ± 15%; (Fig 2B-2D) and respective isotype controls (S2C- S2E Fig). Most importantly, the late hepatic endodermal markers, ALB and AFP were expressed but significantly less higher level when compared with 16TF-iENDO cells (Fig 1F), which was confirmed by immunostaining (S2F-

14TF iENDO cells can be expanded for at least 20 passages without significant loss of their endodermal progenitor phenotype
The rationale for making iENDO cells, rather than direct reprogramming of somatic cells to mature endodermal cells, was to create an expandable population of cells. Therefore 14TF iENDO cells were passaged in endoderm induction medium with 50ng/ml Activin-A and 25 ng/ml Wnt3a for 6 to 8 passages on Matrigel-coated dishes. In addition we expanded the iENDO cells from passage 6/8 to 20 in a less complex and less expensive medium, namely hMAPC medium on 0.1% gelatin-coated dishes (Fig 2A). We reevaluated expression of the mesendoderm, definitive endoderm, epithelial and late endoderm markers after 6-8 and 15-20 passages. Both after 6-8 and 20 passages, most endogenous transcripts remained unchanged compared to day 20 iENDO cells, even if the definitive endoderm transcripts, SOX17, eFOXA2, CXCR4, CKIT, CXCR4, and eFOXA1, as well as the hepatoblast transcript HNF6 were increased, while eHNF4a decreased (Fig 2E-2G). Importantly, more mature pancreatic and hepatocyte transcripts remained very low. In addition, transcript levels of all transgenes were significantly decreased in 6-8 and 20 passage iENDO cells (Fig 2H). These studies demonstrated that it is possible to create a population of endodermal progenitor cells from hMAPCs that can be expanded at least 20 passages, using a relatively non-expensive medium with a stable transcriptional profile. Genomic stability of early (P4) and late passage (P20) 14TFs iENDO shown that there is no significant genome wide aberrations observed (S5A and S5B Fig).

Grafting of 14TF iENDO cells in vivo
To test their in vivo differentiation capacity, we transplanted one million 14TF ENDO cells (passage 6) under the kidney capsule of BALBC Rag2 -/γc -/mice in 30μl of PBS (Fig 3A). After grafting, the weight and glucose levels of the mice were regularly assessed. Three mice were sacrificed after 3 weeks (N = 3) and the other (N = 12) were maintained untill 3 months after grafting. Unfortunately, of the last group, one mouse died in the 12 th week. In all mice, we found that a tumor was formed surrounding the kidney. The tumor was larger at 3 months after transplantation (Fig 3B).
Graft sections were stained with Hematoxylin and Eosin to assess the graft morphology. None of the grafts had invaded into the kidney. In general, at 3 weeks (N = 3) and at 3 months (N = 12), 14TF iENDO cells-derived tumor cells were homogenous containing endodermal cells, akin to a hepatocellular carcinoma (Fig 3C and 3D). However, the graft of the 12 mice sacrificed at 3 months, showed a more heterogeneous morphology, even if most cells still appeared endodermal (Fig 3D).
We immunostained the grafts for the hepatocyte markers AFP, ALB, AAT, HNF4α and the transgene OCT4. Most cells stained positive for OCT4 after 3 weeks (S6A- S6E Fig) (N = 3) and after 3 months (N = 7), cells in the grafts stained positive for the liver markers AFP, ALB  Fig and S10A Fig) and NGN3 (S10A Fig), and very few also for SST (S10B Fig). Insulin-or glucagon-positive cells were not detected. In addition, cells in the graft stained positive for the neuroendocrine marker CHGA (S10C Fig), the paneth cell marker, LYSOZYME (S10D Fig) and the intestinal brush-border microvilli marker, VILLIN-C (S10E Fig). Also, many of the cells still stained positive for OCT4 (S10F Fig). In addition, mesodermal lineage markers, as an example the cardiac specification marker NKX2.5 and ectodermal lineage markers, as an example the neuronal marker β-TUBULIN-III were not observed by immunostaining (S11A and S11B Fig) or by qRT-PCR (S11C and S11D Fig). Control, non-grafted kidney sections did not stain positive for any of these markers.
We also performed qRT-PCR on grafts harvested at 3 weeks (N = 3) and at 3 months (N = 9). At 3 weeks transcript levels for ALB, AFP, AAT and TTR were significantly higher in the graft compared with the initial iENDO cell population, even if mature hepatocyte markers G6PC, NTCP, MRP2 and CYP3A4 were not significantly increased (S12A and S12B Fig). Likewise, after 3 months, transcript levels for ALB, AFP and TTR, but not AAT, G6PC, NTCP, MRP2 and CYP3A4 were significantly increased in the graft compared with the initial cell population (S12A and S12B Fig). Consistent with the immunostaining results, we also detected increased levels of the pancreatic endoderm transcripts PDX1, NGN3, NKX6.1, NEUROD1, PAX4, SST, but not INS or GCG; as well as the gut markers, HNF1β and KRT19, CDX2 in mouse 2 (S12C-S12E Fig). Transcripts for the transgenes were similar before and after grafting, both at 3 weeks and at 3 months (S12F Fig).
We analyzed the mouse serum for presence of human albumin. After 3 weeks, human albumin levels were 1,189 ± 615.44 ng/ml (N = 4), and after 3 months 5,777 ± 7,928 ng/ml (N = 12). In comparison, human blood serum albumine levels are 265,499 ± 12,5354 ng/ml (Fig 3E), while human albumin levels in untransplanted mice were 377± 3.28ng/ml. The levels of human albumin depicted are the levels measured minus the levels found in untranplanted mouse serum. Negligible levels of human C-peptide (2.735 ± 0.06 pmol/L) were detected in the blood of (mouse 1&2), 3 months after grafting (Fig 3F), while no c-peptide was detected in the blood of mice 3 months after grafting. c-Peptide levels in human serum were 518.63 ± 584.62 pmol/L.
These studies demonstrate thus that 14TF iENDO cells can differentiate in vivo to multiple endodermal lineages, including hepatoblast, pancreatic endoderm, as well as gut endoderm like cells.

14TF iENDO cells can differentiate into hepatocyte-like cells
As spontaneous differentiation occurred from iENDO cells to cells with hepatoblast features in vivo, we assessed if iENDO cells could be directly differentiated in vitro to cells with mature hepatocyte features, using methods developed in the lab for hPSCs [40]. As iENDO cells have an endodermal phenotype, we started differentiation from day 8 (hepatic endoderm specification) or day 12 (hepatoblast maturation) of the hPSC protocol onwards (protocol A vs. B, respectively (Fig 4A and S13A Fig). To determine if there are morphological differences between iENDO derived hepatocyte organoids cultured following protocol-A (day 20) or following protocol-B (day 16) and cultured with or without DMSO, pictures were taken at the end of each protocol. Not many differences were observed between organoids of the two protocols. However, organoids cultured with DMSO appeared smaller than the spheroids cultured without DMSO (Fig 4B and S13B Fig). No DMSO condition cultured organoid core seems to have dead cells compare to the DMSO by H&E Staining (Fig 4B and S13B Fig). An additional variable was addition of DMSO, as others [11,15,43,44] and we have demonstrated that this improved the overall differentiation of PSCs to hepatocyte-like cells (HLCs) [40]. Initial studies demonstrated that the highest transcript levels for ALB, AFP, AAT, MRP2, CYP3A4, TTR, G6PC, eHNF4α (Fig 4C) were detected when cells were differentiated using protocol A with addition of DMSO compare to the protocol-B (S13C Fig). Therefore, we chose protocol-A + DMSO for further evaluation and comparison with ESCs differentiated as 3D organoids.
Expression levels of ALB, AFP, AAT, MRP2 and TTR transcripts were ± 1 log lower than d20 differentiated iENDO compared with day 30 3D ESC-HLCs (differentiated as described [45], until day 20 and as organoids until day 30) (Fig 4C). Transcripts for eHNF4α, and eHNF6, CYP3A4 and G6PC in 3D iENDO differentiated organoids were comparable with 3D-ESC-HLCs (Fig 4C). Addition of DMSO to either iENDO or hESC differentiation improved differentiation (Fig 4C). Pancreatic endocrine gene transcript levels were not induced under hepatic differentiation conditions (S14A and S14B Fig). When compared to primary human hepatocytes (PHHs), 3D iENDO and 3D hESC hepatocyte-like cells continued to express AFP, while transcripts for mature hepatocytes were significantly lower compared with PHHs ( Fig 4C). Transgenic tOCT4 transcripts remained detectable in iENDO-HLCs, albeit at decreased levels compared to undifferentiated iENDO cells ( Fig 4C). As negative control, iENDO cells were cultured as organoids without growth factors and small molecules ( Fig  4C and S13C Fig), which did not induce increased expression of hepatocyte transcripts.

14TF iENDO cells can differentiate into pancreatic endocrine cells
We also assessed if 14TF iENDO cells could be differentiated towards endocrine pancreatic cells. 14TF-iENDO cells (passage 6-12) were cultured as organoids with combinations of growth factors and small molecule epigenetic modifiers known to induce endocrine pancreas differentiation as described in (Fig 5A) [18,21,22]. We observed compact spheroids as early as 2 days after plating the cells. The morphology of spheroids on day 12 and day 30 of the protocol is shown (Fig 5B). There was not much difference observed between spheroids on day 12 and day 30, except the size of the organoid which was smaller at day 30 than day 12. On day 12, Live/Dead and H&E staining demonstrated that cells in the organoids were viable (Fig 5C  and 5D). In parallel, hESCs differentiated into pancreatic endocrine cells by using the protocol described in our group previously [18].
On day 12 (stage 1) and day 30 (stage 2) the expression of pancreatic marker genes was assessed by qRT-PCR (Fig 5E), and compared with undifferentiated 14TF-iENDO cells (day 0), hPSCs differentiated to pancreatic endocrine cells at day 30, and primary human islets. We detected a significant increase in gene expression of PTF1A, PDX1, NGN3, NKX6.1, NEU-ROD1, NKX2.2, PAX4, MAFA, MAFB, EHNF6 and SST but not for INS and GCG, nor NKX6.1 (master regulator β-cells) or ARX (master regulator α-cells) (N = 5) (Fig 5E). When compared with hPSC progeny, transcript levels in day 30 iENDO progeny were similar to those of hPSC progeny, except that INS and GCG, ARX, PAX6 expression was significantly higher in hPSC progeny. When compared with primary islets, INS, GCG and SST expression in both iENDO and hESC progeny was significantly lower, whereas endocrine pancreatic genes, including NGN3, were significantly higher expressed in iENDO and ESC progeny (Fig 5E). Also, hepatocyte marker transcripts were not induced in iENDO derived pancreatic progeny (S14C and S14D Fig). Following differentiation of iENDO cells, the OCT4 transgene remained expressed. When iENDO cells were cultured without growth factors and small molecules as organoids, pancreatic transcripts were not induced (Fig 5E).
Day 12 organoids were encapsulated in 2% agarose gel and paraffin embedded sections immunostained for the pancreatic endoderm markers. PDX1 and NGN3 were expressed ( Fig  5F and 5G lower image).
Thus, consistent with the in vivo data, iENDO cells can be guided towards endocrine pancreatic cells in vitro, even though not to mature β-or α-cells.

Discussion
In this study, we demonstrated that hMAPCs can be reprogrammed into a population of iENDO cells using a combination of 14TFs and culture medium containing Activin-A, Wnt3a and BMP4. iENDO cells can subsequently be expanded for at least 20 passages using a Induced endodermal progenitor cells differentiation to hepatocytes and endocrine pancreatic cells relatively inexpensive medium and can differentiate in vivo and in vitro into multiple endodermal cell types.
Reprogramming hMAPCs to iENDO cells was dependent on the presence of OSKM as well as the growth factors Activin-A, Wnt3a and low doses of BMP4. Indeed, removal of either of these components from the reprogramming method failed to induce the generation of epithelioid clusters and induction of endogenic mesendoderm and definitive endoderm transcripts. This is in line with studies described by Funa S et al. demonstrating that interaction between the TF OCT4, the Nodal effectors SMAD2/3, and Wnt effector β-catenin is required to activate primitive streak gene expression by co-operatively binding to the chromatin upstream of primitive streak gene loci [46]. Moreover, Loh et al. demonstrated that low concentrations of BMP4 and Wnt initially specify the anterior primitive streak [47], whereas Li et al. found that BMP4 combined with OSKM can activate epithelial markers such as E-CADHERIN, EPCAM, OCLN (OCCULUDIN) [48]. Recently, Li et al. and Zhu et al. also demonstrated that definitive endoderm-like cells expressing SOX17 and FOXA2 could be generated from mouse and human fibroblasts by transient overexpression of OSKM combined with a cocktail of epigenetic molecules. These endodermal cells could then subsequently be converted to pancreatic β-cells and hepatocytes, even if knockdown of p53 was required to create the mature progeny from human endodermal cells [25,33,49].
We also assessed whether, aside from OSKM, the other 10 TFs used in the 14TF used to reprogram hMAPCs to iENDO cells were required. To identify the required complement of TFs for reprogramming MAPCs into iENDO, we removed single TFs from 14 transcription factors and assessing creation of cells with iENDO features. When we removed OCT4, SOX2, KLF4, SOX17, or MIXL1, GATA4, HNF6 from the 14TF pool, the efficiency of iENDO formation decreased, as the resultant cell population did not express the same levels of endogenous definitive endoderm transcripts and the cells could not be expanded (data not shown).
When iENDO cells were grafted under the kidney capsule of immunocompromised mice, we found differentiation solely to endodermal cell, including cells with hepatoblast and endocrine pancreatic characteristics and to a lesser extent cells with intestinal cell characteristics, but not ectodermal or mesodermal positive cells. The reason why, in one mouse assessed at 3 months, iENDO cells differentiated predominantly to hepatoblast-like cells, and in the other mouse, to pancreatic and gut endoderm is currently not clear. However, the iENDO cells formed progressively increasing tumors, which did not invade the kidney. These data are consistent with a study by Seguin et al. wherein SOX17 was overexpressed in hESCs, which lead to the formation of tumors containing both mesodermal and endodermal cell types [50]. As is the case of iENDO cells, the SOX17 overexpressing hESCs also still expressed OCT4, which we believe may be responsible for the continued proliferation in vivo. However when Cheng et al. showed that when CXCR4/CKIT enriched endodermal progenitors or human foregut stem cells derived from hPSCs were transplanted in mice, formation of mature endodermal tissues or foregut endodermal cells, respectively, was observed without formation of a tumor [51,52]. Live/Dead staining of 3D aggregates at day 12 of differentiation (Live-green, Red-dead) (scale bar 50μm). D) H&E staining of day 12 iENDO pancreatic organoids. E)Relative gene expression (to PPIG, log scale) analysis represented in a heat-map of pancreatic endocrine marker 14TFs iENDO day 0, day 12 iENDO pancreatic organoids, day 30 3D iENDO pancreatic organoids and day 30 hESC pancreatic organoids compared with primary human Islets. All data represent mean of three independent experiments. F) Immunostaining for PDX1 (20X) in 293x transduced PDX1 (top) and day 12 iENDO pancreatic organoids (bottom) (scale bar 100μm, representative of N = 3) with respective isotype controls. F) Immunostaining for NGN3 (20X) in day 12 iENDO pancreatic organoids with respective isotype controls (scale bar 50μm, representative of N = 3). Ã p<0.05, ÃÃ p<0.01 and ÃÃÃ p<0.001 determined by unpaired 2-tailed Student's t-test. https://doi.org/10.1371/journal.pone.0197046.g005 In both studies, OCT4 expression was no longer detected in the PSC-derived endodermal progenitor cells that were grafted.
Of note, the hepatoblast-like cells in the iENDO progeny secreted significant amounts of human albumin. When compared with published reports wherein HLCs derived from hESC or hiPSC, albumin levels in the blood of our mice grafted with iENDO cells (±1200 ng /ml after 3 weeks, and >5000 ng/ml after 3 months) appeared to be similar or even higher, except in the mouse wherein the graft had differentiated towards predominantly pancreatic and gut endoderm cells. . Albumin levels detected in mice grafted with induced hepatocytes (iHEPs) were reported to be as high as 300 μg/ml in the Du et al. study after 7 weeks [24], but only 400 ng/ml in the Huang et al. study [23]. Differences in the number of cells that secrete albumin in vivo as well as the location of the graft, as well as the mouse model used may impact the results observed. Nevertheless, except for the Du et al. studies, albumin levels detected following grafting of iENDO cells are in line with or higher than published results for HLCs derived from PSCs and iHEPs.
In line with the in vivo data demonstrating that iENDO cells can mature further towards different endodermal lineages, we demonstrated in directed in vitro differentiation studies that iENDO cells can generate hepatocyte-like cells and endocrine pancreatic cells. As iENDO cells already have an endoderm/definitive endoderm phenotype, we omitted the initial commitment step (using Activin-A and Wnt3a) commonly used to differentiate PSCs to endodermal cells towards the pancreatic or hepatocyte lineage.
iENDO progeny expressed signficantly higher levesl of numeorus hepatic genes, to levels near to ±1 log lower compared with hESC progeny. However, as has been described [19,20], the differentiation system did not induce fully mature hepatocytes, as the iENDO progeny, like hESC-progeny, did not express mature hepatocyte transcripts (such as G6PC and CYP3A4), did not have significant CYP3A4 activity and continued to express high levels of AFP. As has been demonstrated by others[55, 56] and us [40] for hPSC differentiations to HLCs, addition of DMSO increased HLCs differentiation of iENDO cells.
To differentiate iENDO cells towards endocrine pancreatic cells we used a protocol adapted from Pagliuca et al. [21] and Alireza Rezania et al. [22], wherein was demonstrated that mature β-cells can be generated from hPSCs. Although pancreatic endocrine transcripts were induced by the combination of growth factors and small molecules described, and organoid formation, the master regulators for α-cell (ARX and PAX6) and β-cell (NKX6.1) remained very low, and consistent expression of INS and GCG transcripts also remained very low. When compared with hESC differentiation as we described previously [18], we found that iENDO cells differentiated less robustly. hESCs progeny expressed higher levels of the master regulators for α-cells, which was in accordance with higher transcript levels for GCG. Moreover, INS transcripts were significantly higher in hESCs progeny compared with iENDO progeny. A possible explanation for the lower degree of maturation of iENDO-derived progeny compared with hESCs progeny could be the persistent expression of the transgene OCT4 after differentiation. When compared with primary islets, iENDO cells were significantly less mature, with retained expression of NGN3, and have significantly lower levels of INS and GCG transcripts as well as their upstream transcriptional regulators. Of note, NGN3 transcripts in hESCs progeny were also still significantly higher than in mature islets.
In conclusion, we demonstrated that it is possible to reprogram hMAPCs that do not have a robust endodermal differentiation potential to a long-term expandable population of endodermal progenitor cells using TFs and growth factors. These iENDO cells can differentiate both in vitro and in vivo to multiple different more mature endodermal cell types, even if fully mature hepatocytes and β-cells could not be generated and persistent expression of the pluripotency TF, OCT4, may be responsible, at the same time, for the continued proliferation of the iENDO cells in vivo, and in part for the incomplete maturation of the iENDO cell to mature endoderm cell types. Future studies wherein OCT4 is only transiently expressed may shed light on this. Nevertheless, the current study describes an alternative cell source, which might be a starting point for the creation of mature endodermal cells that could be considered for regenerative medicine and drug studies.