Turning Human Epidermis Into Pancreatic Endoderm

1 Center of Regenerative Medicine in Barcelona, Dr. Aiguader 88, 08003 Barcelona, Spain. 2 Julia McFarlane Diabetes Research Centre, Department of Microbiology and Infectious Diseases, and Institute of Inflammation, Infection and Immunity, Faculty of Medicine, The University of Calgary, 3330 Hospital Dr. N.W., Calgary, Alberta T2N 4N1, Canada. 3 Networking Center of Biomedical Research in Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN). 4 Institució Catalana de Recerca i Estudis Avançats (ICREA). 5 Gene Expression Laboratory, Salk Institute for Biological Studies, 10010 North Torrey Pines Rd., La Jolla, California 92037, USA. * On sabbatical leave. Address correspondence to: Pere Santamaria, e-mail: psantama@ucalgary.ca


Introduction
t present, the transplantation of pancreatic islets, or whole pancreas from a cadaveric donor, are the only alternatives to insulin therapy, to restore normoglycemia in type 1 diabetic patients. Islet transplantation can afford insulin-independence to some patients, and can improve the quality of life of others, even when insu-lin independence is not achieved [1]. However, islet transplantation is not currently the standard treatment for type 1 diabetes. Possibly one day, the development of immunosuppressive strategies capable of selectively blunting autoreactivity, without impairing systemic immunity, coupled with improved islet isolation and preservation techniques, may justify a more general application of this approach. However, the realization of this dream will only be possible if the problem of limited organ availability is eventually solved.
Differentiation of human embryonic stem (hES) cells into pancreatic β-cells has been long considered as a possible solution to this problem. ES cells, derived from the inner cell mass of preimplantation embryos, can both be propagated indefinitely without loss of the undifferentiated state, and prompted to differentiate into cells belonging to endoderm, mesoderm, or ectoderm lineages. Early studies showed that mouse ES cells could be differentiated in vitro into cells displaying phenotypic and morphological properties of islet cells [2][3][4][5][6]. More recently, it has been shown that hES cells can be differentiated into pancreatic endoderm and, much less efficiently, into clusters containing insulin-producing cells [7][8][9][10][11][12][13].
The need to use human embryos to generate hES cells as a source of transplantation material poses significant ethical concerns. The discovery that mouse fibroblasts can be reprogrammed into induced pluripotent stem (iPS) cells by transduction of four key transcription factors (Oct4, Sox2, Klf4, and c-Myc) provided a potential solution to this ethical issue, and brought forth the notion that derivation of tissues from patient-specific stem cells is possible [14]. These studies were replicated by others [15], and extended to the generation of human iPS cells from fibroblasts [16][17][18]. However, it has not been addressed whether iPS cells prompted to differentiate into pancreatic endoderm progressively upregulate markers of differentiation with similar kinetics/rate as ES cells.
The purpose of this study was to determine the feasibility of differentiating pancreatic endoderm from human iPS cells. More specifically, we examined iPS cells generated from human skin keratinocytes, as these are significantly more amenable to iPS generation than fibroblasts. Also, we monitored the kinetics of marker expressions at different stages of differentiation.

Immunoflourescence and flow cytometry
EBs and monolayers were fixed in 4% PFA for 2 h at 4ºC, washed 3 times with PBS, and treated with Target Retrieval Solution, pH 9 (Dako) for 1h at 60ºC and 30 min at RT. Samples were washed 3 times with Tris-buffered saline (TBS), and incubated with block buffer (TBS, 0.5% Triton X-100, 6% donkey serum) for 2 h at RT, followed by overnight storage at 4º C, and 2 h at RT. Subsequently,

Results and discussion
The pancreas arises from the definitive endoderm, a specialized epithelium that forms the primitive gut tube, and specifies the structures that differentiate into the organs arising from this germ layer. Dorsal and ventral buds originating from the gut tube fuse to form the posterior foregut. Factors produced by the notochord, such as activin A and FGF7, coupled with retinoic acid (RA) signaling and inhibition of sonic hedgehog signaling, specify expression of the key transcrip-tion factor Pdx1 by the dorsal pancreatic endoderm [21,22]. The latter gives rise to all the endocrine, exocrine, and ductal cells of the pancreas. Specific sets of transcription factors are expressed at each of these stages. The definitive endoderm, for example, expresses Sox17 and FoxA2. As it progressively differentiates into the primitive gut tube, posterior foregut, and pancreatic endoderm, it acquires expression of Hnf1β and Hnf4α, followed by Pdx1, HlxB9, and Hnf6, and ultimately, Ngn3, Pax4, Nkx6.1, Nkx2.1, and Isl1, among others [22].
We first attempted to differentiate the ES4 cell line using Kroon's protocol [13], but with modifications [9]. Undifferentiated ES4 cells expressed the stem cell factors Sox2 and Oct4 (Figure 1), and lacked expression of ectoderm, mesoderm, and endoderm markers (not shown). Also, they were invariably negative for endocrine markers, or markers of definitive or pancreatic endoderm, including Hnf4α, Hnf1β, Sox17, Pdx1, C-peptide, and glucagon ( Figure 1). On day 10 of differentiation, the monolayers were sliced into fragments, detached, and cultured for 3-4 days in ultra-low attachment plates to promote formation of EB-like structures (Figure 2A Figure 2B). However, most of the Sox17 + and FoxA2 + cells contained in individual EBs coexpressed Hnf4α and Hnf1β ( Figure 2B). These markers were found in the primitive gut tube, which gave rise to the pancreas and other endoderm organ primordia [21]. Furthermore, a significant fraction of FoxA2 + cells also expressed Pdx1, Nkx2.2, Isl1, or Nkx6.1 ( Figure 2B). These factors are markers of committed pancreatic epithelium, and play key roles in endocrine cell differentiation [22]. In fact, a small number of cells in these structures were C-peptide-or glucagonimmunoreactive ( Figure 2B), suggesting a capacity to differentiate into endocrine cells. We next asked if this protocol could generate pancreatic endoderm from iPS cells reprogrammed from human keratinocytes [20]. For this purpose, we used the KiPS3F1 line generated by reprogramming human foreskin keratinocytes through retroviral transduction with Oct4, Sox2, and Klf4. These keratinocytes have been extensively characterized and shown to be fully reprogrammable [20]. Also, they appear indistinguishable from hES cells regarding colony morphology, growth properties, expression of pluripotencyassociated transcription factors and surface markers, global expression profiling, as wells as in vitro and in vivo differentiation potential [20]. The data shown below were obtained with KiPS3F1 cells, but similar results were obtained with KiPS4F1 cells [20], which represent a pool of ~20 independent KiPS colonies generated with Oct4, Sox2, Klf4, and c-Myc. As shown in Figure 3A, KiPS cells co-expressed Sox2 and Oct4, in addition to other markers, which are typically expressed by hES cells cultured under identical conditions, such as alkaline phosphatase, SSEA3, and SSEA4 ( [20] and data not shown). Upon differentiation and mechanical processing, the monolayers and resulting EB structures were morphologically similar to those obtained with differentiated ES4 cells ( Figure 3B). Nuclei of KiPS cells did not stain for Sox2 and Oct4 ( Figure 4A). The resulting EBs contained significant, but variable portions of FoxA2 + /Sox17 + cells, such as we have seen for EBs derived from  ES4 cells (Figures 4B and 5). Most of the KiPSderived EB cells displaying high nuclear content for FoxA2 or Sox17 co-expressed Hnf4α and Hnf1β, Pdx1, Nkx2.2, and Isl1 ( Figure 5). Many of the hES and KiPS-derived EB cells that did not express these markers, contained nuclei that stained for Sox17, suggesting that they belonged to a less differentiated endoderm state (see Figure 2B, second row from top, and Figure 5, first and third rows from top). As in the case of ES4-derived EBs, a very small fraction of cells were C-peptideand/or glucagon-immunoreactive ( Figure 5).
We determined whether there were any substantial differences in the ability of ES4 and KiPS cells to progressively differentiate into pancreatic endoderm in vitro. For this purpose, we measured changes in marker expression by flow cytometry and/or Taqman real-time PCR at the following different stages: -day 1 (after culture in sodium butyrate, Wnt3a, and activin A) -day 3 (after culture in activin A and 0.2% FBS) -day 7 (after culture in FGF7 and 0.2% FBS) -day 10 (after culture in KAAD, RA, and noggin) -day 13 (EB-like structures cultured in DMEM supplemented with 1% B27).
As shown in Figure 6, ES4 and KiPS cells underwent remarkably similar changes in gene expression in response to these in vitro developmental cues. Culture in the presence of activin A and Wnt3a induced a marked upregulation of the mix1 homeobox-like 1 gene (Mixl1), a primitive streak marker, and T (brachyury) and Gsc, which are expressed in the mesendoderm ( Figure 6A). This was associated with a slight upregulation of the endoderm-specific marker FoxA2 ( Figure 6A). Removal of Wnt3a and addition of FBS led to further upregulation of FoxA2 and appearance of Hnf4α, a primitive gut tube marker, as well as Hex and Isl1, a pancreatic endoderm marker, in these Mixl1 + /T + /Gsc + cells.
To determine the extent of the culture conditions leading to the generation of definitive endoderm cell types, we compared undifferentiated ES4 and KiPS cells, and the corresponding day 3  differentiated cells for reactivity to FoxA2-and Sox17-specific antibodies by flow cytometry. As shown in Figure 6B, most day 3 cells, but not their undifferentiated precursors, were FoxA2 + /Sox17 + , consistent with efficient differentiation into definitive endoderm. From days 3-7 (withdrawal of activin A and exposure to FGF7), the cultures downregulated Gsc and upregulated Hnf4α, Isl1, Pdx1, and Slc2A (Glut2), while maintaining similar lev-els of Mixl1, T, and FoxA2. By day 10 (in response to RA, noggin, and inhibition of hedgehog signaling), the cells had downregulated Mixl1, T, and Gsc to the levels seen in undifferentiated cells. Also, they expressed maximal or near-maximal levels of FoxA2, Hnf4α, Isl1, Pdx1, and Slc2A, which are markers of committed pancreatic epithelium, and began to express Gcg. Whereas, we could not detect significant changes in insulin mRNA expression (not shown), presumably because the percentage of C-peptide-expressing cells was very low, as shown above ( Figure 5). These data are consistent with a previously reported role for RA signaling in facilitating the conversion of mesendoderm into pancreatic endoderm [10]. Furthermore, they accord with the immunofluorescence microscopy analyses shown above, and demonstrate that ES4 and KiPS cells respond in a remarkably similar fashion to the in vitro developmental cues explored herein. Several terminally-differentiated cell types other than fibroblasts, including mouse B lymphocytes [23], and liver and gastric cells [24], have been reprogrammed to iPS cells by transduction with Oct4, Sox2, Klf4, and c-Myc. We have recently shown that human keratinocytes, derived from the epidermis, are amenable to reprogramming into iPS cells by transduction with Oct4, Sox2, and Klf4, with or without c-Myc. In fact, we provided evidence that keratinocytes are more amenable to reprogramming than other cell types. This may be because they display a transcriptional profile significantly closer to hES cells than the profile seen in fibroblasts [20]. In our hands, keratinocytes typically yield iPS cell colonies within 10 days after retroviral transduction, as compared to more than 20 days for fibroblasts [17]. Already at this early stage of reprogramming, a significant fraction of the cells within these colonies expressed several hES cell markers, including alkaline phosphatase, Tra-1-81, SSEA4, Sox2, and Oct4, and lacked expression of keratinocyte markers, such as keratin 14. Fur- thermore, the overall reprogramming efficiency of keratinocytes approached 1%, which is ~100-fold of that reported for fibroblasts [20,25].
It is still not clear whether iPS cells derived from all cell types can differentiate into pancreatic endoderm, or insulin-producing islet cells. The fact that iPS cells derived from fibroblasts can do so [26], should not be taken as proof that iPS cells derived from other cell types are equally amenable to differentiation into these structures. Likewise, it is not clear to what extent differentiation of iPS cells into pancreatic endoderm mimics the differentiation of hES cells into pancreatic endoderm. In the present study, we have shown for the first time that keratinocyte-derived iPS cells can be used to generate pancreatic endoderm in vitro. We found that the efficiency of pancreas endoderm formation is comparable to that seen with conventional hES cells (albeit lower than that reported for CyT203 hES cells) [13]. Furthermore, we have shown that both hES and iPS cells undergo remarkably similar changes in the expression of several different differentiation markers. We have not determined whether iPS-derived pancreatic endoderm can differentiate into mature, glucoseresponsive β-cells in vivo. However, the presence of a few C-peptide and glucagon immunoreactive cells in the endoderm structures derived from both hES and iPS cells, suggests that the β-celldifferentiation potential of hES and KiPS cells is similar. These observations, coupled with the reported ability of pancreatic endoderm grafts to differentiate into islet cells in vivo [13], suggest that it may one day be possible to transplant diabetic patients with grafts derived from their own skin cells.