Biliary Tree Stem Cells, Precursors to Pancreatic Committed Progenitors: Evidence for Possible Life-long Pancreatic Organogenesis

Abstract

Peribiliary glands (PBGs) in bile duct walls, and pancreatic duct glands (PDGs) associated with pancreatic ducts, in humans of all ages, contain a continuous, ramifying network of cells in overlapping maturational lineages. We show that proximal (PBGs)-to-distal (PDGs) maturational lineages start near the duodenum with cells expressing markers of pluripotency (NANOG,OCT4,SOX2), proliferation (Ki67), self-replication (SALL4), and early hepato-pancreatic commitment (SOX9,SOX17,PDX1,LGR5), transitioning to PDG cells with no expression of pluripotency or self-replication markers, maintenance of pancreatic genes (PDX1), and expression of markers of pancreatic endocrine maturation (NGN3,MUC6,insulin). Radial-axis lineages start in PBGs near the ducts’ fibromuscular layers with stem cells and end at the ducts’ lumens with cells devoid of stem cell traits and positive for pancreatic endocrine genes.

Biliary tree-derived cells behaved as stem cells in culture under expansion conditions, culture plastic and serum-free Kubota’s Medium, proliferating for months as undifferentiated cells, whereas pancreas-derived cells underwent only ∼8-10 divisions, then partially differentiated towards an islet fate. Biliary tree-derived cells proved precursors of pancreas’ committed progenitors. Both could be driven by 3-dimensional conditions, islet-derived matrix components and a serum-free, hormonally defined medium for an islet fate (HDM-P), to form spheroids with ultrastructural, electrophysiological and functional characteristics of neoislets, including glucose regulatability. Implantation of these neoislets into epididymal fat pads of immuno-compromised mice, chemically rendered diabetic, resulted in secretion of human C-peptide, regulatable by glucose, and able to alleviate hyperglycemia in hosts. The biliary tree-derived stem cells and their connections to pancreatic committed progenitors constitute a biological framework for life-long pancreatic organogenesis.

Introduction

The global incidence of diabetes mellitus has increased dramatically over the past few years and continues to rise. The quest for curative therapies that normalize blood glucose levels and provide independence from exogenous insulin therapies impacts patients with type 1 diabetes (T1D) and a significant subset of patients with type 2 diabetes (T2D) who have a functional deficiency in insulin production. Islet transplantation is viewed as an ideal treatment for such patients, but it is constrained by the limited yields of quality donor pancreata that can be utilized to isolate islets¹. The hope has been to identify one or more precursor populations that can be lineage restricted to islet cells and, thereby, constitute a nearly limitless and reproducible supply of transplantable and functional islets².

Determined stem cells for pancreatic cell therapies have not been considered an option based on evidence that there are no or only rare pancreatic stem cells in postnatal tissues³. The few studies in which OCT4⁺ and SOX2⁺ multipotent stem cells have been identified in adult pancreas have indicated also their rarity^4-6. Instead, the postnatal pancreas has long been thought to contain only committed progenitors, found in pancreatic ducts^{7, 8} and, more recently, in pancreatic duct glands (PDGs) by Thayer and associates⁹. These precursors are reported to be limited in their proliferative and self-renewal potential.

The phenotype of these progenitors and their actual contribution to the endocrine compartment are actively debated¹⁰. Signs of human beta-cell replication and expression of beta-cell markers in pancreatic ductal structures have been described in situations such as pregnancy¹¹ or with underlying inflammation (e.g. pancreatitis, T1D), and rejection of pancreatic grafts^12-15, though the biological relevance of these phenomena to the maintenance of functional beta-cell mass throughout life remains to be elucidated. Regeneration of beta-cells in postnatal pancreas is mediated primarily by beta-cells³ except for experimental conditions under which sub-total beta-cell ablation occurs resulting in plasticity of other pancreatic cells that are able to become beta-cells^16-18.

Recently, a new source of islet precursors has been identified in biliary trees in donors of all ages^{19, 20}. They comprise multiple subpopulations of determined stem cells with indefinite expansion potential in culture and that can mature to hepatocytes, cholangiocytes or islets depending on the microenvironment in vitro or in vivo¹⁹. In subsequent studies it was found that the peribiliary glands (PBGs), the stem cell niches of the biliary tree, connect directly into the Canals of Hering, the intra-hepatic stem cell niches, making a continuous network of stem cells contributing to the formation of liver and biliary tree²¹.

We now provide evidence that the biliary tree is also a reservoir of stem cells for the pancreas. The biliary tree and the pancreatic ducts also comprise a continuous, ramifying network connecting the pancreas to the duodenum at two sites: the minor papilla, the connection of the dorsal pancreatic duct to the duodenum, and the major papilla, the ampulla of Vater, the connection to the duodenum by the hepato-pancreatic common duct, the merged ventral pancreatic duct and common bile duct. The hepato-pancreatic ampulla (Figs 1A, S1 and S2) opens into the duodenum, where it releases the bile from the liver and the pancreatic exocrine contents. Inter-individual anatomical variations exist (Fig S2), as the pancreatic and bile duct can be fused into a hepato-pancreatic duct of various lengths, and can be separated by an interposed fibromuscular layer. The biliary tree also has peribiliary glands that can be within the duct walls, intramural glands, or ones extending by a connection from the duct surface, the extramural glands²². The roles of extramural peribiliary glands are unknown.

Figure 1. Panel 1. The peribiliary glands and pancreatic duct glands form a continuous network withinthe biliary tree and pancreatic ducts.

Panel 1. The hepato-pancreatic common duct. The biliary tree and its peribiliary glands (PBGs) are part of a continuous network connecting to the pancreatic duct and its pancreatic duct glands (PDGs). (A) The schematic drawing shows the bile duct, the main pancreatic duct and their fusion at the hepato-pancreatic ampulla (the papilla of Vater). The sectioning planes indicate the regions from which the proximal (B) and distal (C) histological sections were taken. (B) A section in the proximal plane shows the histology of the bile duct and pancreatic duct in the immediate proximity of the fusion in the hepato-pancreatic ampulla. Just before the formation of the hepato-pancreatic ampulla, the fibromuscular layer, dividing bile and pancreatic ducts, is mostly absent; PBGs and PDGs are connected and indistinguishable from each other. The dashed lines indicate the PBGs (in green) versus the PDGs (in red). Note the striking similarities in the histological appearance of bile duct versus pancreatic duct and of PBGs versus PDGs. Hypercellular foci are marked with asterisks*. (C) A section in a distal plane shows the histological appearance of PBGs and PDGs in the separated ducts, divided by a thick fibromuscular layer (FM layer). (D) A plane intermediate between B and C shows glands intermingled and crossing the FM layer (artificially shown as a green area). Note the evidence of ductal and alveolar cuboidal and columnar cells, eosinophilic acinar-like cells and mucinous cells in the PBGs and PDGs. Several hypercellular foci are present (marked with asterisks*) in the PBGs of the hepato-pancreatic common duct, suggesting intense proliferative activity at the site. (E) Higher magnification of a hypercellular focus from the boxed area in D. Cells in hypercellular foci display a high nucleus/cytoplasm ratio and are tightly packed. (F) A compartment of the PDGs can be identified by Mucin 6 (MUC6) staining. The PDG cells show PDX-1 staining. Alveolar and ductal spaces are filled with amylase. For more detailed schematics of the hepato-pancreatic common duct see Figs S1 and S2. Scale bar=250 μm

Analyses of gene expression profiles indicate progressive maturational lineages of cells with stem cell traits within PBGs to ones with committed progenitor traits within PDGs. To define stemness, we used markers of pluripotency (OCT4,SOX2,NANOG) found in embryonic stem (ES) cells and found requisite for reprogramming of somatic cells to become induced pluripotent stem (iPS) cells ^23-27. Further discussion on the background of these and other markers used for characterizing the cells within the biliary tree and pancreatic duct systems is given in the online supplement. Complementing the in situ findings, we provide evidence that biliary tree-derived cells behave as stem cells in culture and are precursors to committed pancreatic progenitors similar to those in PDGs. In summary, we present evidence to suggest the biliary tree and pancreatic networks are connected anatomically and functionally to comprise maturational lineages relevant to pancreatic organogenesis.

Results

A Ramifying Network of Stem Cell and Progenitor Cell Niches in the Biliary Tree and Pancreas

The biliary tree, the pancreatic ducts and their associated glands, PBGs and PDGs, demonstrate striking similarities histologically (Fig 1, Panel 1). At the hepato-pancreatic common duct, the region of the merger of the ventral pancreatic duct and common bile duct, large numbers of glands can be found, some of which are intermingled into the fibromuscular tissue. Those in the hepato-pancreatic common duct are continuous with ones associated with the bile duct and with the pancreatic duct. In the immediate proximity of the fusion between the pancreatic and bile duct, glands crossing the interposed fibromuscular layer can be observed (Fig 1-Panel 1D, Figs S1 and S2).

The glands throughout the network harbor a plethora of cell types, including ductal and alveolar cuboidal and columnar cells, eosinophilic acinar-like cells, and mucinous cells. Hypercellular foci (noted with asterisks) are observed frequently in the cells of the hepato-pancreatic common duct, suggesting an intense proliferative activity at the site (Fig 1, Panel 1, B-E) but not in those of the pancreatic ducts.

In Fig 1, Panel 2, we show that multiple pluripotency genes (OCT4A, SOX 2) and an endodermal transcription factor (SOX17) are co-expressed in the nuclei of cells in the PBGs within the hepato-pancreatic common duct. Immunohistochemical studies of the PBG’s cells confirm the presence of the pluripotency genes and also positivity for LGR5, a marker associated with endodermal stem cells. In the lower row of images, we show that the pluripotency genes are absent in the cells of the PDGs. LGR5 is also absent (data not shown). By contrast, NGN3 is expressed in the PDG cells found associated with pancreatic ducts nearest to the duodenum (Fig 1, Panel 2).

Figure 1. Panel 2. Stem Cells are present in PBGs versus committed progenitors inPDGs. See also Figs.2-3, S3 and S7.

(G-J) Immunofluorescent stainings show the co-expression of SOX17, OCT4A and SOX2 in the cells of PBGs. (K-M) Immunohistochemistry also show that OCT4A, SOX2 and LGR5 are evident in the cells in PBGs. (N2P) By contrast, only rare or no cells in PDGs were found expressing OCT4A or SOX2; none were found expressing LGR5 (data not shown); and a significant proportion of the PDG cells expressed the transcription factor, NGN3, a marker of endocrine progenitor cells. Scale bar = 50 μm.

Phenotypic Transitions from PBGs (Proximal) to PDGs (Distal) implicate Transitions from Stem Cells to Committed Progenitors

More detailed analyses of the phenotypic traits implicate two separate but overlapping maturational lineages identified by gradients in gene expression. One is a proximal (PBGs)-to-distal (PDGs) axis of maturation from the duodenum and extending into the pancreatic ducts. The other is a radial axis of maturation starting at the fibromuscular layers within duct walls and extending to cells at the ducts’ lumens.

We provide evidence of the proximal-to-distal maturational lineage axis starting in PBGs in the hepato-pancreatic common duct near to the duodenum, transitioning to pancreatic ducts, thence to PDGs and finally to mature or maturing pancreatic islet cells (Figs 1-3, S3 and S7). Cells expressing markers of pluripotency (NANOG, OCT4, SOX2), self-replicative ability (SALL4) (Fig S7), and of hepato-pancreatic endodermal commitment (SOX9, SOX17, PDX1, LGR5) are present in the PBGs of the normal, adult hepato-pancreatic common duct (Figs 1-3, S3 and S7). In separate studies, we have shown that they also express CD133 ²¹. Approximately 9% of the PBG cells co-express all of them, and of the remainder, from 5 to 30% express at least one of them; none expressed NGN3 (Fig S3). By contrast, cells in the PDGs had no cells co-expressing pluripotency genes; none expressed NANOG or SALL4; fewer than 5% expressed either Oct4A or SOX2, and this expression was cytoplasmic; all expressed PDX1; and most expressed NGN3 (Figs 1-3, and S3).

Figure 3. Panel 1 (A-H). The maturational lineage gradients in terms of endocrine traits.

Panel 1 (A-H) shows a proximal (PBGs)-to distal (PDGs) gradient in expression of early-to-late pancreatic commitment markers. PDX1 and SOX 17, transcription factors of early hepato-pancreatic commitment are co-expressed in nuclei of PBG cells (A). By contrast, SOX 17 is lost, and only PDX1 is expressed in the nuclei of PDG cells. (E) MUC6 shows an affinity for a differentiated compartment of the glandular epithelium; it is found in a portion of the cells in the PBGs (B) and in all cells in PDGs (F). NGN3, a marker of endocrine committed progenitors, is not found in the nuclei of any cells of the PBGs (B) but in a large proportion of cells within PDGs (F). EpCAM is found in only a subset of the cells in the PBGs (C) but in almost all of the cells of the PDGs (G). See also Fig S6. Insulin can be observed in rare cells in PBGs (C) but is found in a large number of cells in PDGs (G). A high proliferative activity is found in PBGs, as indicated by Ki-67 staining (D), whereas Ki-67 positivity is rarely found in PDGs (H). Scale Bars = 100 μm.

In parallel to the findings with the pluripotency genes and with SALL4, we found that cells within PBGs demonstrated co-expression of SOX17 and PDX1 (Fig 3A), but expression only of PDX1 in the PDGs (Fig 3E). There was no expression of NGN3 in PBGs (Fig 3B), but it was strongly expressed in PDGs (Fig 3F). Insulin was found only in rare cells in PBGs (Fig 3C) but was present in large numbers of cells in PDGs (Fig 3G). There was strong evidence of proliferation (Ki67) cells in PBGs (Fig 3D) but not in PDGs (Fig 3H).

The radial-axis maturational lineage consists of stem cells in the PBGs deep within the walls of the hepato-pancreatic common duct and near the fibromuscular layers (Fig 3I-L). The PBGs near these fibromuscular layers contained cells that did not express EpCAM, NGN3, insulin or any other islet hormone but co-expressed, within the nuclei, the pluripotency genes (NANOG, OCT4, SOX2), SALL4 (Fig S7) and the endodermal commitment genes (SOX17, PDX1, LGR5) (Figs 1-3). In a separate study, we found that they also expressed SOX9 and CD133²¹. As one progresses towards the luminal surface of the duct, the expression of the pluripotency genes and SOX17 faded and, in parallel, there was maintenance of PDX1 along with appearance of and then increasing expression of EpCAM and insulin (Fig 3I-L).

Figure 3. Panel 2 (I-L).

Panel 2 (I-L) shows a radial axis maturational lineage extending from the fibromuscular layer within the duct walls to the lumens of the ducts. (I) A radial maturational lineage process begins near the fibromuscular layer and ends at the luminal surface of the both bile ducts and pancreatic ducts. (J) shows the magnified image from (I). Both show double immunofluorescence for EpCAM and Insulin. Insulin⁺ cells can be observed interspersed or in aggregates among glandular and ductal cells in the hepato-pancreatic ampulla, and undergoing commitment toward pancreatic endocrine fates. Insulin+ cells are mostly EpCAM+ (red arrows). EpCAM+/insulin- cells are also present (green arrow). Especially near the fibromuscular layer, the EpCAM- cells can be observed (white arrow) and that are negative also for insulin. (K) Immunofluorescence for Insulin in adult pancreas. Pancreatic islets are insulin positive and represent the positive control for the staining. Notably, cells of interlobular pancreatic duct are mostly insulin negative (white arrow). (L) Immunohistochemistry for EpCAM (brown) counterstained with PAS (pink) in hepato-pancreatic common duct. A radial maturational lineage can be observed also for maturation towards acinar cell. EpCAM-/PAS+ cells (pink arrows) are located at the luminal surface; EpCAM+/PAS+ cells (brown arrows) at the middle; and EpCAM-/PAS- cells (black arrow) very near the fibromuscular layer. Scale bars= 100 μm.

The net sum of these transitions in the radial-axis and proximal-to-distal axis lineages was a shift from stem cells in the PBGs to committed pancreatic progenitors in the PDGs. The committed progenitors had little evidence of stem cell traits or proliferation but increasing evidence of pancreatic endocrine differentiation. We have not yet done many studies on acinar cell differentiation but assume that it must occur also in a similar fashion to that for islets. The findings of committed progenitors within pancreatic ducts and in PDGs corroborates those of many prior investigators^{7, 9, 18, 28, 29}. A summary of the transitions in gene expression is given in Table S5.

EpCAM, an Intermediate Marker in the Lineages

EpCAM proved to be an intermediate marker in the radial and in the proximal-to-distal axis maturational lineages. In the in situ studies, the PBGs nearest to the fibromuscular layers had no expression of EpCAM (Fig 3I-L); EpCAM⁺ cells appeared in PBGs at levels nearing to or at the lumens of the ducts. By contrast, EpCAM⁺ cells were evident in all of the PDGs. Flow cytometric analyses of cell suspensions indicated that EpCAM⁺ cells comprise 2.4%-3.8% of human fetal pancreata between 16-20 weeks gestation (Fig S6A). Immunofluorescence staining of the human fetal pancreata showed that EpCAM⁺ cells located around pancreatic ducts co-express SOX9, Ki67 (Fig S6B). They also expressed PDX1 and NGN3 (data not shown). EpCAM co-expressed with endocrine cell markers such as insulin, C-peptide and glucagon within the PDGs and into the islets (Figs 3 and S6C).

In Vitro Evidence of Stem Cell Populations within the Biliary Tree versus Committed Progenitors within the Pancreas

Culture selection for stem cells and/or committed progenitors occurred under conditions comprised of tissue culture plastic plates and serum-free Kubota’s Medium (KM). This medium has been shown previously to select for early endodermal stem/progenitors, and with minor modifications works well also for mesenchymal cell stem/progenitors^{19, 30, 31}. Here we corroborate our prior findings ¹⁹. The human biliary tree stem cells (hBTSCs) formed colonies that expanded readily on plastic and in KM, generating colonies of cells dividing initially every ∼40 hours and then slowing to a division every 2-3 days. Two types of colonies of small cells (7-9 μm in diameter) were observed from fetal, neonatal, pediatric and adult human biliary tree tissue (Fig 4A-C). Type one colonies were comprised of cells with an undulating, swirling morphology (“dancing cells”) and that initially did not express EpCAM. With time in culture, the cells acquired EpCAM expression at the edges of the colonies in parallel with a slight increase in cell size (10-12 μm) and of markers indicating slight differentiation. The type two colonies were comprised of cells that expressed EpCAM on every cell and formed “carpet” like colonies of tightly packed, uniformly cuboidal-shaped cells. Both type I and type II cell populations expressed NCAM, CD133, CD44, Cytokeratins (CK8, 18 and 19), PDX1 and SOX17, but did not express NGN3, nor mature pancreatic islet or liver markers (e.g. insulin, albumin), complementing the findings published previously^19-21. See also Figs S4 and S5. Both types of colonies of hBTSCs persisted as undifferentiated cells for months as long as the cells were maintained in KM. Evidence of their division capacity was that by 8 weeks, the cells routinely had gone through at least 19-25 divisions and generated colonies of more than 500,000 cells, all derived from 2-3 cells, corroborating past findings¹⁹. When cultured in KM, hBTSCs retained expression of pluripotency markers such as OCT4, SOX2 and NANOG (Fig S4A-C), whereas they did not display the endocrine committed progenitor markers such as NGN3 (Fig S5A). Cultures of hBTSCs did not show markers of mature pancreatic islet cells, though rare clusters of cells were found displaying insulin and nuclear MAFA positivity (Fig S5). If subjected to a serum-free, hormonally defined medium designed to drive the cells to a pancreatic islet fate (HDM2P), the cells lost expression of the pluripotency markers and acquired expression of NGN3, insulin, MUC6 and MAFA (Figs S4 and S5). Thus, cultured hBTSCs resembled their in vivo counterparts and in HDM2P gave rise to cells with properties overlapping with those of committed progenitors (compare in situ studies in Figs 1-3 with culture ones in Figs 4, S4, and S5).

Figure 4. Comparison of biliary tree cells versus pancreatic cells under expansion culture conditions.

The expansion conditions are culture plastic and serum-free Kubota’s Medium. Both biliary tree tissue and fetal and adult pancreas yield colonies initiated by a small number of cells. Biliary tree stem cell (hBTSC) colonies (A-C) divide initially about every 36-40 hours, and then slow to a division every 2-3 days, continuing to expand indefinitely for months and yielding large colonies, each containing over 500,000 cells by ∼8 weeks of culture. They vary in morphology from flattened, monolayer colonies to ones that are slightly 3-dimensional. Two types of colonies have been observed: those in which the colonies are EpCAM negative (B) but give rise to EpCAM⁺ cells at the edges, and those in which every cell, from the outset, expresses EpCAM (C). As noted in the phase image in A, the type I colonies are connected to the type 2 suggesting a precursor-descendent relationship. This was confirmed by the fact that all cells become EpCAM⁺ if inducers of differentiation are added. This indicates that the type I cells are giving rise to the type 2 cells. The cells from fetal pancreas yield colonies that behave as committed progenitors (D-K). If plated on culture plastic and in Kubota’s Medium (D, E, H), the cells are EpCAM⁺ from the outset and form colonies that initially look similar to those of hBTSCs, but they are essentially negative for C-peptide (H). Yet even when on culture plastic and in serum-free Kubota’s medium, they go through only 8-10 divisions and then begin to aggregate (F and G). They also become NGN3+ (Fig S5) that is accompanied by partial endocrine differentiation with expression of other endocrine markers such as C-peptide (I), glucagon (J) and somatostatin (K). Magnification: A-E (10X); F-K (20X). Scale Bars = 100 μm.

By contrast to the findings with hBTSCs, the colonies of pancreatic cells, derived from fetal or adult tissue, plated on plastic and in KM, behaved as committed progenitors. The pancreas-derived cells were, from the outset, negative for SOX17, uniformly expressed EpCAM (Fig 4H), and morphologically were similar to the colonies of hBTSCs (compare Figs 4A, D and E). Yet the colonies of pancreatic cells underwent a total of ∼8-10 divisions with division rates slowing by the second week, then stopping in proliferation altogether by the end of the second week or early in the third week. In parallel, the cells underwent aggregation and then detachment to form floating spheroids (Fig 4F and G). The spheroids contained cells that transitioned steadily to have increasing expression of mature endocrine makers such as C-peptide, glucagon (GCG), and somatostatin (SST). This occurred even when these cells were maintained on plastic and in serum-free KM (Fig 4J-K). These phenomena occurred with fetal and adult pancreas, but the number of cell divisions observed was maximum (∼10-12) with the fetal pancreas cultures.

Differentiation Conditions induced both Biliary Tree and Pancreas-derived Cells to Spheroid Formation with Ultrastructural, Electrophysiological and Functional Features Typical of Pancreatic Islets

The conversion to neoislet-like spheroids occurred most rapidly (in a few days) and most robustly when cells were transferred to HDM-P along with embedding them into a 3-dimensional hydrogel comprised of 60% type IV collagen/laminin (1:1 ratio) and 40% hyaluronans (Fig 5A and B). Hematoxylin/eosin staining of spheroids showed cord-like structures (Fig 5C). The differentiation towards an islet fate was faster and stronger than the one observed in HDM-P alone, but it was still partial. We assume that other factors are required in the culture conditions to elicit full maturation in vitro. Immunofluorescent staining showed a slight increase in C-peptide positivity along with expression of glucagon and CK19 (Fig 5D-G). The immunohistochemistry data were corroborated by RT-PCR findings of fetal pancreatic cultures in KM versus HDM-P (Fig 5H). The fetal pancreatic cells cultured in KM showed a pattern of gene expression consistent with a committed progenitor state with high levels of expression of CXCR4, EpCAM, HNF3b, HNF6, Prox1, HB9 and PDX1. When shifted to HDM-P, there was a decrease in expression of progenitor markers in combination with an increase in the endocrine commitment markers such as NGN3, NEUROD-1, PAX6, ISL1, GCG, Insulin, and SST.

Figure 5. Formation of neoislet-like spheroids paralleled by endocrine differentiation.

(A-B) Formation of neoislet-like spheroids occurs within a few days if cells are cultured in serum-free, hormonally defined medium designed to drive the cells towards a pancreatic islet fate(HDM-P) and embedded into hydrogels consisting of 40% hyaluronans and 60% type IV collagen/laminin mixture (1:1 ratio). (C) Hematoxylin/eosin staining of a neoislet demonstrates cord-like structures. Immunohistochemistry of cells at intermediate and late stages indicate that cells become positive for mature islet markers, including C-peptide (C-pep), glucagon (GCG), and somatostatin (SST) (D-G). (H) RT-PCR analyses corroborate the immunohistochemistry in a survey of genes expressed when cells are in Kubota’s Medium versus in HDM-P. See also online supplement Figs S4 and S5. Scale bars are as labeled on the different images.

Ultrastructural analyses of both the pancreatic progenitor cell colonies (Fig 6A) and the pancreatic colony-derived spheroids (Fig 6B-D) showed distinct features. The former showed cells in small size, a high nucleus/cytoplasmic ratio, and cells tightly compacted together, while the latter showed typical characters for immature pancreatic islets. This interpretation is corroborated by the presence of vesicles containing insulin, proven by immuno-electron-microscopy for insulin in the granules. Undifferentiated cells released negligible levels of C-peptide, whereas differentiated cells increased significantly their C-peptide release (Fig 6E). Moreover, the release was found to be regulatable by high glucose (16.7mM D-glucose) or 100 μM tolbutamide (TOL). The neoislets, but not the undifferentiated cells, were electro-physiologically responsive (Fig 6F).

Figure 6. Functional assays for the differentiated pancreatic progenitor cells.

(A-D) Transmission electron microscopy (TEM) and immune-electron microscopy of pancreatic progenitor cells that derive from differentiation of cells to spheroids in culture. (A) The progenitor cells are tightly compacted (similar to findings with hBTSCs) and have a high nucleus to cytoplasmic ratio. (B-D) After differentiation, the cell interactions become looser, and the cells grow in size. The differentiated cells contain secretory granules, large amount of mitochondria and well-arranged endoplasmic reticulum generated in the cytoplasm. The granules proved positive for insulin in differentiated neoislets; the white arrows indicate gold particles bound to insulin. (E). Stimulated C-peptide secretion assay shows that undifferentiated progenitor cells (Undiff) secrete negligible amounts of human C-peptide, whereas differentiated neoislets released significantly higher amounts of C-peptide (significance levels equal to 0.01 to 0.001). The C-peptide released is regulatable as observed after incubation with Low (5.5 mM) versus High (16.7 mM) glucose concentrations or 100 μM TOL. (F) Electrophysiological recordings show electrical responsiveness in differentiated neoislet cells (Diff), but not in the undifferentiated (Undiff) pancreatic progenitor cells. Scale bars= 1 μm

In vivo Transplantation of Neoislets ameliorates diabetes in streptozotocin (STZ)-treated mice

In order to verify the functional differentiation in an in vivo setting, neoislets generated in culture from pancreatic progenitor cells were transplanted into the epididymal fat pads of immuno-deficient Rag^-/-/Ilrg2r^-/-mice subsequently rendered diabetic with STZ. Monitoring of body weight (Fig 7A) and non-fasting blood glucose levels (Fig 7B) in transplanted experimental (Exp) mice versus controls (Ctrl) showed improvement in mice transplanted with the neoislets. All of the mice that received neoislets lived longer than 120 days. The blood glucose levels in mice treated with STZ increased gradually and reached above 600mg/dl. Some mice made diabetic by STZ, but not transplanted, died by post-operative day (POD) 60. Some of those were able to survive longer only because of receiving daily long-acting insulin. Human C-peptide was detected in the serum of transplanted mice at low levels on POD 37 and at higher levels on POD 60. The levels in vivo were responsive to IP glucose administration (Fig 7C). IPGTT showed glucose tolerance in transplanted mice is improved over that in the controls (data not shown).

Figure 7. Transplantation of neoislets can alleviate hyperglycemia in diabetic mice.

(A-B) Representative body weight and non-fasting glucose levels in transplanted versus control mice indicate improvement in transplanted mice. The body weight decreased, while the blood glucose level increased gradually in all of the control mice. In the control group, 4 out of 7 mice died at or near post- operative day (POD) 60, when their blood glucose levels were above 600mg/dl for more than 20 days. The rest of the mice survived longer than 120 days by using the long-term insulin treatments. Body weight increased steadily, while the blood glucose level decreased and then persisted at ∼150-280mg/dl in the transplanted mice. In the transplanted mice, the blood glucose levels increased only after the epididymal grafts were explanted (see arrows). (C) Serum C-peptide was detected in transplanted mice at low levels on POD 37 and at higher levels on POD 60. On POD 60, the levels were also responsive to glucose administration (2g/kg weight) via IP. (D) A macroscopic view of the explanted epididymal fat pads two months after transplantation. Note the residual grey sutures remaining at the top of each of the testes. Hematoxylin and Eosin staining of the explanted epididymal fat pads show that the transplanted cells either fused into the fat pads loosely or grew as a solid mass (E-F). The graft area is well-vascularized. Immunofluorescence staining shows high levels of expression of human C-peptide (arrows) in the transplanted sites (G). Magnification, E-G (10X); H (20X). Scale bars are as labeled on the images.

The fat pads in the group of Exp versus Ctrl mice were surgically removed and examined macroscopically and histologically (Fig 7D and G). Hematoxylin/eosin staining of them showed that transplanted neoislets integrated within the tissue growing as a solid mass with neovascular structures (arrows in Fig 7F). Immunofluorescence staining of the tissue showed high levels of expression of human C-peptidein the transplanted cells (Fig 7G and H).

Discussion

The biliary tree constitutes the stem cell reservoir for the pancreas. That realization is striking given the years of studies contributing evidence that stem cells are not present postnatally in the pancreas. This past history of investigations has indicated that the pancreas is essentially devoid of stem cells, and this has fueled efforts to identify ES cells, iPS cells, or determined stem cells from non-endodermal sources and that might be lineage restricted to islets for treatment of diabetic patients fate^33-37. The ES or iPS cells have been studied most extensively, and the findings have led to considerable achievements in lineage restriction to pancreatic islet fate^33-37. Yet the strategies are compromised by low efficiency and high costs with only a small fraction of the cells completing the process and requiring transplantation for several months in vivo to achieve a reasonable extent of maturity to functional islets. In addition, there remains a risk of ∼15% teratoma formation for the ES or iPS cells that do not complete the commitment/differentiation process³⁸, a concern that has resulted in major efforts to find ways of eliminating this problem^{39, 40}.

Transdifferentiation of mesenchymal stem cells (MSCs) from bone marrow, adipose tissue, cord blood, amniotic fluid ^41-46 or from amniotic epithelia⁴⁷, is an alternative being actively investigated⁴⁸ with the distinct advantage that these precursors do not have the risk of tumorigenic potential. However, lineage restriction of these precursors to an islet fate is even less efficient, necessitating prolonged culture under defined conditions and/or transduction of transcription factors for pancreatic endocrine commitment (e.g. PDX1, NGN3). Such requirements are challenging for the use of the resulting islets in clinical programs. Moreover, success to date has resulted in cells that are hybrids of islet cell and mesenchymal cell phenotypes, a fact of unknown significance.

More recently, transdifferentiation of pancreatic acinar cells to islets was demonstrated following transfection of at least 3 key endocrine transcription factor genes (PDX1, NGN3 and MAFA) ^{49, 50}. These are exciting findings but not yet relevant to aspirations for clinical programs given safety issues regarding transfection of cells ⁵¹.

We show here that determined stem cell populations, present throughout life, are precursors for pancreatic committed progenitors in the pancreatic duct glands, and are present in the ramifying, continuous network of ducts and associated glands of the biliary tree. The evidence for stemness in PBGs is extremely strong both functionally from their extraordinary proliferative capacity and multipotency in culture and also given the co-expression in the nuclei of the same cell of multiple genes classically associated pluripotency as demonstrated in ES cells or used to reprogram somatic cells to iPS cells.

Counter arguments derive from ongoing controversies regarding OCT4A in adult tissues, especially cancers, in which its expression has sometimes been found to be from pseudogenes with distinct functions relative to those demonstrated in ES or iPS cells^52-55. Yet the presence of so many pluripotency genes in nuclei of individual biliary tree cells; the evidence of large numbers of these cells with pluripotency gene expression; and the evidence of down-regulation of these genes during differentiation in culture trumps the dismissive arguments and supports our interpretation that they indicate that subpopulations of the PBG cells are stem cells. Future studies will clarify the functionality of OCT4A in cells of the biliary tree.

Our findings on the connections of the biliary tree’s stem cell niches, the PBGs, to the reservoirs of pancreatic committed progenitors, the PDGs, complement our prior ones showing that the PBGs connect to the niches of the intra-hepatic stem cells, the canals of Hering²¹, providing a cellular infrastructure and maturational lineages for ongoing organogenesis of both liver and pancreas postnatally. That network developed embryologically from outgrowths, or anlage, from the mid-gut endoderm, and gave rise to the liver, biliary tree, and pancreas that are then connected to the duodenum²⁰. Schematics of the anatomical connections are shown in Figs S1 and S2.

Rare stem cell populations have been identified within the pancreas and that express pluripotency genes (OCT4A, SOX2) ^{5, 29, 56}. Zhou et al ⁵⁶ stated that they found only 1-30 to 1-200 per 100,000 of these cells. The rarity of these stem cell populations in adult pancreas is in striking contrast to the relatively high numbers (0.5-2%) of stem cells that we have shown exist in the Canals of Hering in fetal and adult liver^{30, 57}, of comparable numbers in most of the biliary tree³², and even higher numbers (5-9%) found by us in the PBGs of the hepato-pancreatic common duct. Comparable findings were made for tissue from donors of all ages from fetuses to geriatric adults (see Tables S3 and S4, listing information on the donors).

The hBTSCs were comprised of at least two distinct cellular subpopulations: those that were EpCAM-negative and that become EpCAM⁺ with differentiation, and those that were EpCAM⁺in situ and when freshly isolated. The in situ and in vitro investigations indicate that EpCAM-negative cells are precursors to EpCAM⁺ cells. The highest numbers of EpCAM-negative cells were in PBGs near the fibromuscular layer deep within the walls of the hepato-pancreatic common duct. Our findings here and in previous reports indicate that these cells express SOX9, PDX1, SOX17, LGR5, and CD133, along with the pluripotency genes (OCT4, SOX2, NANOG, SALL4) and have the highest levels of Ki67/PCNA ^19,21. By contrast, EpCAM⁺ cells were located in PBGs near to or at the lumen of the hepato-pancreatic common duct. EpCAM positivity was evident on most, if not all, of the cells around the ducts of the PDGs in both fetal and adult tissues. The percentage of EpCAM⁺ cells (2-4%) in fetal pancreas by flow cytometric analyses was in sharp contrast to the percentage found in fetal liver (>80%). EpCAM⁺ cells in pancreas have lower or no levels of Ki67, no expression of NANOG or SALL4, and only rare cells with cytoplasmic, not nuclear, expression of SOX2 and OCT4A. These EpCAM⁺ cells in the pancreas have high levels of NGN3, PDX1, MUC6, and CK19. Therefore, EpCAM expression in the pancreas is an indicator of cells in transition to or that have already become committed progenitors.

The connections between biliary tree and pancreas have been suggested in a number of prior investigations summarized in a recent review ²⁰. Among these are those by Slack and associates⁵⁸, who demonstrated that there are insulin⁺ cells in the biliary tree of murine hosts, and studies by Nakanuma and his associates on anatomy, biology and pathology of the biliary tree²². Wells and associates demonstrated that SOX17 is a molecular “toggle” switch that in one position yields biliary tree and, when inactivated, yields pancreas ⁵⁹. The molecular mechanisms are not fully understood but involve connections with the jagged-notch-HES pathway, a signaling pathway able to be driven by FGF10 ⁶⁰. Knockout of the HES gene also results in formation of endocrine functions within the biliary tree⁶¹. This suggests that achievement of greater expansion of cells able to give rise to neoislets could be achievable by manipulations of these signaling pathways.

Confirmation of the in situ findings regarding relationships between hBTSCs and pancreatic committed progenitors occurred with cultures of the tissues. Cells from the two tissues reacted quite distinctly to expansion conditions comprised of culture plastic and KM, developed originally for clonogenic expansion of hepatoblasts and then found successful for expansion of multiple types of endodermal stem cells and progenitors ^{19, 30, 62}. The medium is comprised of any basal medium with low calcium (<0.5mM), no copper, and supplemented only with insulin, transferrin/Fe, a mixture of free fatty acids bound to purified albumin, high density lipoprotein, and prepared in low oxygen. Mature liver parenchymal cells and mature pancreatic cells do not expand or even survive in this medium. Cultures of biliary tree tissue, prepared from any portion of the biliary tree, and in Kubota’s Medium yielded colonies of cells that remained proliferative and undifferentiated for months, whereas those from fetal or adult pancreas, under the same condition, expanded only transiently, went into growth arrest, formed aggregates that subsequently detached from the dishes and, in parallel underwent partial endocrine differentiation. Implicit in these findings is that these are committed progenitors.

Numerous efforts have been made to optimize expansion of pancreatic precursors. A recent report summarized the findings of assays with more than 23 different culture conditions with human fetal pancreas to identify ones optimal for cell expansion. Yet under the best conditions identified, the pancreas cells, even if fetal, underwent only ∼10-12 divisions^{63, 64}. A recent report identified conditions that enabled pancreatic duct cells to achieve up to 18 divisions⁶⁵. In another report, investigators modified human fetal pancreatic progenitors with telomerase reverse transcriptase (hTERT) to enhance cell replication without loss of stem/progenitor cell properties and also introduced PDX1 into these cells to promote them to differentiate into insulin-expressing cells⁶⁶. In the adult, islet beta-cells are assumed to regenerate primarily from pre-existing islet beta-cells³. However, more recent findings indicate that insulin is expressed by early progenitors and cannot be used to define just the mature or terminally differentiated cells⁶⁷. Stanger and associates⁶⁸ have concluded that specification to a pancreatic fate results in the loss of proliferative potential. We agree and conclude that the differences in expansion potential of the pancreatic cells versus those of the biliary tree are inherently in their genetic and cellular “blueprint”, a “blueprint” of commitment, not just a missing environmental condition.

Though biliary tree and pancreatic cells have dramatically distinct proliferative potential, they are both able to form functional neoislets in an appropriate microenvironment tailored to mimic the islet extracellular matrix chemistry in vivo and with soluble signals known to drive islet differentiation. The neoislet formation in culture occurred with rapid kinetics in approximately a week in these conditions. It is noteworthy that lineage restriction of ES and iPS cells or MSCs to an islet fate requires sequential treatments with sets of growth factors and matrix components and/or transfection with key transcription factors (e.g. PDX1), in protocols that lasts for 4 to 6 weeks^{33, 34}. By contrast, the hBTSCs are already at ∼stage 4 (and their descendants, the committed pancreatic progenitors, even further along) of the 5 known stages of the stepwise differentiation process for islet formation described previously⁶⁹. They are poised to generate islets rapidly (in days to a week) in an appropriate microenvironment and without the need for genetic manipulations.

The neoislet-like spheroids that resulted from the differentiation conditions were functional in vitro and in vivo. The transplantation of cells pre-differentiated in culture to neoislets ameliorated the diabetic condition of STZ-treated mice. Comparable findings were obtained in prior studies with transplantation of neoislets from hBTSCs into diabetic hosts^19,19. These findings are consistent with the interpretation that hBTSCs are precursors to pancreatic committed progenitors.

All treatments of the diabetic condition occurred in mice with neoislets derived from cells differentiated in culture and transplanted into the epididymal fat pads. The transplanted neoislets generated C-peptide-positive cell clusters in the fat pads. Amelioration of hyperglycemia occurred within 2 months after transplantation, a period of time that is significantly shorter than those reported in transplants of ES- or iPS-derived cells (3-4 months)^{33, 34} due, it is assumed, to the fact that the biliary tree-derived and pancreas-derived cells are so much further along the differentiation pathway towards an islet fate. The C-peptide positivity in vivo from the transplanted neoislets derived from pancreatic cells occurred faster (30-60 days) and with a stronger (100-350 mg/dL) insulin response as compared with neoislets derived from hBTSCs under the same conditions corroborating our findings in prior studies¹⁹; the neoislets from hBTSCs, prepared under the same conditions as those for the pancreatic committed progenitors, yielded later responses (POD days 60-90 days) and with a weaker response (50-70 mg/dL) of insulin. We interpret this to mean that the hBTSCs are precursors to the committed progenitors within the pancreas, an interpretation corroborated by our in situ analyses of the maturational lineages. It is important now to learn how the hBTSCs and their descendants, pancreatic committed progenitors, respond to pancreatic injuries and to learn the details and mechanisms of regenerative responses.

The radial axis and proximal-to-distal axis in the maturational lineages in biliary tree and pancreas have parallels to those in the intestine. The radial axis of maturation within bile ducts is similar to that of the intestinal stem cells in the crypts progressing to the mature cells at the tops of the villi. The intestine’s proximal-to-distal axis is indicated by the changes in the phenotypes of the mature cells along the length of the intestine (e.g. from esophagus to stomach to small intestine to large intestine).

The hBTSCs offer a novel way to generate neoislets for use in clinical programs. Correction of forms of diabetes with severe insulin deficit is successful with islet transplants, but the strategy is dependent on access to large numbers of organs from cadaveric tissue. The ready availability of biliary tree tissue from existing liver and pancreatic transplantation programs or from gallbladder surgeries, and the extensive expansion potential of the hBTSCs in culture under wholly defined conditions enable these stem cells to be a viable option for clinical programs in the treatment of diabetes and other pancreatic diseases.

Materials and Methods

Tissue Sourcing

Adult human biliary tissues were dissected from intact livers and pancreases obtained but not used for transplantation into a patient. They were obtained through organ donation programs via United Network for Organ Sharing (UNOS). Those used for these studies were considered normal with no evidence of disease processes. Informed consent was obtained from next of kin for use of the tissues for research purposes, protocols received Institutional Review Board approval, and processing was compliant with Good Manufacturing Practice.

In addition, we received human fetal biliary tree and pancreata from an accredited agency (Advanced Biological Resources, San Francisco, CA) from fetuses between 16-20 weeks gestational age obtained by elective pregnancy terminations. The research protocols were reviewed and approved by the Institutional Review Board for Human Research Studies at the University of North Carolina at Chapel Hill. Further details on the samples are given in Tables S3 and S4.

The data given in all the figures and the pictures represent the findings in at least three independent experiments. For the rest of the methods, please see the online supplement.

Figure 2. Panel 1 (A-C). The maturational gradients in gene expression of pluripotency genes.

Panel 1 (A-C) Pluripotency related transcription factors (OCT4, NANOG, SOX2) were strongly expressed in the nuclei of PBG cells. (D-F) By contrast, in PDGs, there was no expression at all of NANOG, and expression of OCT4 or SOX 2 occurred only in the cytoplasm of rare cells. The presence or absence and the subcellular localization of these transcription factors represent specific stages in lineage commitment. Scale bars= 100 μm.

Figure 2. Panel 2 (G-J).

Sections of PBGs in the hepato-pancreatic common duct were triple stained for SOX2 (green), OCT4A (red) and NANOG (blue). The merged image indicates cells co-expressing the 3 pluripotency genes (white). (K-N) Images from PDGs showing rare cells with cytoplasmic, but not nuclear, staining of SOX2 and OCT4A and lacking altogether any NANOG expression. For lower magnification images and a table with quantitation of the numbers of cells with the specific phenotypic properties, see Fig. S3. Scale bars = 100 μm