Multipotent Stem/Progenitor Cells in Human Biliary Tree Give Rise to Hepatocytes, Cholangiocytes, and

Vincenzo Cardinale,1,2* Yunfang Wang,1* Guido Carpino,3 Cai-Bin Cui,4 Manuela Gatto,2 Massimo Rossi,5 Pasquale Bartolomeo Berloco,5 Alfredo Cantafora,2 Eliane Wauthier,1 Mark E. Furth,6 Luca Inverardi,7 Juan Dominguez-Bendala,7 Camillo Ricordi,7 David Gerber,4 Eugenio Gaudio,3*,† Domenico Alvaro,2*,† and Lola Reid1,6*,†

Multipotent stem/progenitors are present in peribiliary glands of extrahepatic biliary trees from humans of all ages and in high numbers in hepato-pancreatic common duct, cystic duct, and hilum. They express endodermal transcription factors (e.g., Sox9, SOX17, FOXA2, PDX1, HES1, NGN3, PROX1) intranuclearly, stem/progenitor surface markers (EpCAM, NCAM, CD133, CXCR4), and sometimes weakly adult liver, bile duct, and pancreatic genes (albumin, cystic fibrosis transmembrane conductance regulator [CFTR], and ). They clonogenically expand on plastic and in serum-free medium, tailored for endodermal progenitors, remaining phenotypically stable as undifferentiated cells for months with a cell division initially every 36 hours and slowing to one every 2-3 days. Transfer into distinct culture conditions, each comprised of a specific mix of hormones and matrix components, yields either cords of hepatocytes (express albumin, CYP3A4, and transferrin), branching ducts of cholangiocytes (expressing anion exchanger-2-AE2 and CFTR), or regulatable C-peptide secreting neoislet-like clusters (expressing glucagon, insulin) and accompanied by changes in correlating with the adult fate. Transplantation into quiescent livers of immunocompromised mice results in functional human hepatocytes and cholangiocytes, whereas if into fat pads of streptozocin-induced diabetic mice, results in functional islets secreting glucose-regulatable human C-peptide. Conclusion: The phenotypes and availability from all age donors suggest that these stem/ progenitors have considerable potential for regenerative therapies of liver, bile duct, and pancreatic diseases including diabetes. (HEPATOLOGY 2011;54:2159-2172)

he extrahepatic biliary tree contains a system of biliary glands (PBGs) are tubulo-alveolar glands found branching ducts connecting the liver to the within the duct walls.3 The glands communicate with Tintestine and plays a vital role in the passage of the bile duct lumens through channels opening bile from liver to gut with the gallbladder operating as into diverticula that occur with regularity around the an overflow compartment and a site for removal of mucosal surface. water, resulting in concentration of bile.1,2 The ventral Stem cells and progenitors have been identified and is connected to the gut by way of the hepato- isolated from livers of all donor ages.4-6 They can be pancreatic common duct, shared with the liver. Peri- culture selected with a serum-free, hormonally defined

Abbreviations: AFP, a-fetoprotein; ASMA, a-smooth muscle actin; CTFR, cystic fibrosis transmembrane conductance regulator; ES, embryonic stem cells; HDM, hormonally defined medium; hHpSCs, human hepatic stem cells; iSP, induced pluripotent stem cell; KM, Kubota’s medium; MSC, mesenchymal stem cell; PBG, peribiliary gland; RT-PCR, reverse-transcription polymerase chain reaction; VEGFr, vascular endothelial cell growth factor . From the 1Department of Cell and Molecular Physiology, Biomedical Engineering, Program in Molecular Biology and Biotechnology, UNC School of Medicine, Chapel Hill, NC, USA; 2Division of Gastroenterology, University Sapienza of Rome, Division of Gastroenterology, Department of Scienze e Biotecnologie Medico- Chirurgiche, Fondazione Eleonora Lorillard Spencer Cenci, Polo Pontino, Rome, Italy, Department of Clinical Medicine, Polo Pontino, Rome, Italy; 3Department of Health Sciences, University of Rome ‘‘Foro Italico’’, Rome, Italy; 4Department of Surgery, UNC School of Medicine, Chapel Hill, NC, USA; 5‘‘Paride Stefanini’’ Department of General Surgery and Organ Transplantation, Sapienza University of Rome, Rome, Italy; 6Wake Forest Institute of Regenerative Medicine, Winston Salem, NC, USA; and 7Diabetes Research Institute, University of Miami, Miami, FL, USA. Received May 14, 2011; Revised July 14, 2011; accepted July 20, 2011.

2159 2160 CARDINALE ET AL. HEPATOLOGY, December 2011 medium, Kubota’s medium (KM), supportive of Materials and Methods hepatic progenitors but not of mature cells7 and can be driven to adult fates by specific mixes of systemic The materials and methods can be found in the and paracrine signals8 and/or by biomatrix scaffolds.9 online Supporting Information. By contrast, numerous studies claim that there are no Results stem cells but only committed progenitors within adult 10,11 pancreas. Peribiliary Glands Are Sites for Stem/Progenitors Another source of progenitors is in the biliary tree. Within the Biliary Tree. Hematoxylin and eosin stain- It was reported recently that gallbladder epithelial cells ing (Fig. 1) of different regions of the biliary tree shows 12 can differentiate into hepatocyte-like cells and that that peribiliary glands are found throughout the biliary regeneration of extrahepatic bile ducts occurs with a bio- tree but not in the gall bladder. Quantitative assessments absorbable polymer tube within 11 weeks after surgical of the numbers of peribiliary glands and their sizes indi- 13 removal of the common bile duct in pigs. Also, it was cate that the highest numbers are in the hepato-pancreatic b shown that extrahepatic bile ducts in mice have - common duct, and also in branching points in the biliary 14 cells, with secretory granules that are immunoreactive tree such as the cystic duct and common hepatic duct at for insulin and that exhibit glucose-stimulated insulin the hilum. The percentages (within parentheses) are cal- b secretion. Histological studies indicate that the -cells culated as surface area occupied by all PBGs contained in form directly from the bile duct in late the specimen (duct wall)/total area of the specimen. embryogenesis. Connections between biliary tree, liver, Immunohistochemistry and reverse-transcription po- and pancreas have been made evident most recently by lymerase chain reaction (RT-PCR) (Fig. 2; Supporting reports that SOX17 is a molecular ‘‘toggle’’ switch driv- Fig. S2) on tissue samples (in situ) and from primary ing pancreas formation in one direction and the biliary cultures of biliary tree tissue (Figs. 2, 3) show that 15 tree in another and that SOX9-positive cells can be there are cell populations expressing classic endodermal lineage-traced genetically in intestine, liver, and pan- transcription factors (SOX17, SOX9, FOXA2, HNF6, 16 creas. Other investigations implicating the existence of PROX1, SALL4) and surface markers found on endo- common progenitors within the biliary tree for liver and dermal progenitors (CD326/EpCAM, CD56/NCAM, pancreas are summarized in a recent review (Cardinale CD133, CXCR4). The biliary tree stem/progenitors et al., submitted). Our studies corroborate and comple- expressed no or low levels of lineage markers of the ment those prior findings and clarify further that the liver (a-fetoprotein [AFP], albumin, gamma-glutamyl- biliary tree, even in adults, is replete with multipotent transpeptidase [GGT]) and endocrine pancreas (insu- cells that are stem cells, committed progenitors, or a lin, glucagon). Although not all of these markers are mixture of these and able to lineage restrict to differenti- unique to endoderm the constellation is strongly char- ated cells within liver, bile duct, and pancreas. acteristic, enabling us to hypothesize that the biliary

Support: (UNC) Funding derived from a grant from the North Carolina Biotechnology Center (NCBC), GigaCyte Biotech (Branford, CT), Vesta Therapeutics (Bethesda, MD), and from NIH grants (AA014243, IP30-DK065933), NIDDK Grant (DK34987), and an NCI grant (CA016086); (Sapienza University) Dr. Cardinale received salary support from a scholarship from Sapienza University of Rome for the studies that he did at UNC. D. Alvaro and V. Cardinale were supported by FIRB grant no. RBAP10Z7FS_004; D. Alvaro, V. Cardinale, E. Gaudio, and G. Carpino were supported by a grant from Agenzia Regionale Del Lazio Per I Trapianti E Le Patologie Connesse; E. Gaudio was supported by MIUR grants: PRIN#2007, prot. 2007HPT7BA_001 and Federate Athenaeum funds from the University Sapienza of Rome; (Diabetes Research Institute) The studies were funded by grants from NIH, the Juvenile Diabetes Research Foundation, ADA, and the Diabetes Research Institute Foundation. Patent: A patent on the biliary tree stem cells was filed in November, 2009 and is jointly owned by UNC in Chapel Hill, NC, and Sapienza University in Rome, Italy. *These authors contributed equally to the study. †Coequal senior authors. Current address for Yunfang Wang: Stem Cell and Regenerative Medicine Lab, Beijing Institute of Transfusion Medicine, Beijing, PR China, 100850. Address reprint requests to: Lola M. Reid, Department of Cell and Molecular Physiology, RM 34 Glaxo, UNC School of Medicine, Chapel Hill, NC 27599; [email protected]; fax: 919-966-6112 or Domenico Alvaro, Division of Gastroenterology, University Sapienza of Rome, Division of Gastroenterology, Department of Scienze e Biotecnologie Medico-Chirurgiche, Fondazione Eleonora Lorillard Spencer Cenci, Polo Pontino, Rome, Italy; [email protected], fax: 011 39 06 4453319. Copyright VC 2011 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.24590 Potential conflict of interest: Dr. Reid consults, received grants from, and holds intellectuals property rights for Vesta and GigaCyte. She also received grants from and holds intellectual property rights for Vertex. Additional Supporting Information may be found in the online version of this article. HEPATOLOGY, Vol. 54, No. 6, 2011 CARDINALE ET AL. 2161

Fig. 1. Distribution and characterization of PBGs in the extrahepatic biliary tree. (A) Histological images of PGBs in various regions of the bili- ary tree. (B) Density of PBGs, expressed as surface occupied by PBGs acini/total area as evaluated by imaging analysis; the number and cir- cumference were histologically analyzed in the different sites. The hepato-pancreatic ampulla showed the highest density and number of PBGs; roughly equal numbers were found in cystic duct and hilum; fewer were found in bile duct; and none in gallbladder. The percentages are calcu- lated as surface area occupied by all PBGs contained in the specimen (duct wall)/the total area of the specimen). Magnification 10. See also online Supporting Figs. S2-S4. Immunohistochemistry of PBGs in situ shows that PBGs are positive for CK7, CK19, NCAM, CD133, EpCAM, SOX9, SOX 17, and PDX1 but negative (or very low levels) for albumin and insulin. Magnification 40. Online Supporting Figure showing repre- sentative RT-PCR assays show for adult hepatic hilum indicating a broad repertoire of endodermal transcription factors (SOX9, SOX17, FOXA2, PDX1, NGN3, etc.) and classic stem cell surface markers (e.g., EpCAM, NCAM, CXCR4, CD133). tree contains either a common stem cell and/or a col- and also found effective for human hepatic stem cells lection of committed progenitors for liver, bile duct, (hHpSCs).5 Mature epithelial cells of liver, biliary tree, and pancreas. Until these options are defined, we refer and pancreas do not survive in this medium. Colonies to the cells as ‘‘stem/progenitors.’’ There are sugges- of clonogenically expanding cells were observed in tions of a maturational lineage process from peribiliary cultures from all portions of the biliary tree, but glands deep within the bile ducts and ending at the higher numbers of colonies occurred in cultures from duct lumen. With progression toward the luminal sur- cystic duct, hilum, and hepato-pancreatic ampulla face, there is a decease or loss of stem cell or progeni- (Figs. 2, 3, S6, S7). The colonies formed in the selec- tor markers in parallel with acquisition of mature tion conditions were phenotypically heterogeneous, markers of liver (e.g., albumin) in the portion of the with the centers of the colonies having smaller, biliary tree close to the liver (Fig. S5). more undifferentiated cells, and the edges of the Cultures of Biliary Tree Tissue Results in Selection colonies comprised of slightly more differentiated for Stem/Progenitors. Cell suspensions prepared from cells, including ones qualifying to be committed different regions of the biliary tree were plated onto progenitors. culture plastic and in KM, a serum-free, hormonally There were three types of colonies identified, arbi- defined medium (HDM) designed for hepatoblasts17 trarily named types 1-3. Cells in colony type I formed 2162 CARDINALE ET AL. HEPATOLOGY, December 2011

Fig. 2. The presence of early transcription factors (A) such as SOX17 and PDX1 occurs in peribiliary gland cells both in situ and in vitro.In Supporting Fig. S5 is shown the progression in changes of the gene expression with progression toward the luminal surface. There is a decease or loss of PDX1 expression, in parallel with acquisition of mature markers of liver (e.g., albumin). The early transcription factors (e.g., PDX1 and SOX17) expressed strongly in the nucleus in the stem cell colonies cultured in KM on plastic; magnification 20 in (A) and 10 in (B). (B) RT- PCR assays indicate the expression of diverse genes in the colonies from cystic duct versus from gall bladder. The gene expressions from the two tissues are quite similar, but those from cystic duct have weak expression of both albumin and insulin, an expression pattern not found in cells from the gallbladder. spheroids that grew slowly with divisions occurring colonies and with doubling times similar to those in every 3-4 days (Figs. 3, S6). Cells in the centers of type type 2 colonies. Cells at the colony edges expressed 2 colonies were small (7-9 lm), densely packed, uni- EpCAM and either did not express endodermal tran- form with high nucleus to cytoplasmic ratios (Figs. 2, 3, scription factors (e.g., SOX17) or these transcription S7), and phenotypically essentially identical to those of factors were perinuclear; those in the colony centers intrahepatic hHpSCs. They doubled initially every 36- expressed minimal, if any, EpCAM and yet contained 40 hours but slowed to a division every 2-3 days by 4 strong expression of transcription factors both within weeks in culture. Key features of the colony types 1 and the nuclei and/or perinuclearly. Figure 2 shows RT-PCR 2 are that 100% of the cells expressed EpCAM, NCAM, assays comparing the expression of early endodermal CXCR4, CD133 and were negative for AFP and for transcription factors (e.g., SOX17, HNF6, HES1, markers of mature cell types. Cells in type 3 colonies PDX1, NGN3, SALL4), and surface markers (e.g., consisted of flattened, swirling cells with phenotypic EpCAM, CXCR4) and mature cell markers for colonies traits distinct at the edges versus in the middle of the from cystic duct versus gall bladder. HEPATOLOGY, Vol. 54, No. 6, 2011 CARDINALE ET AL. 2163

Fig. 3. Dominant types of stem/progenitor cell colonies from adult biliary trees. Most of the colonies were one of two types. (in Supporting Fig. S6 are images from a third category). (A,B) Small, round, and tight cells in type 1 colonies have a high nucleus-to-cytoplasmic ratio and have phenotypic markers closely similar to those of hHpSCs and with 100% of the cells expressing EpCAM, NCAM, and none expressing AFP. (C,D) The type 2 colonies are comprised of undulating, swirling cells with EpCAM expression at the edges but not interiors of the colonies and with high levels of expression of SOX17, PDX1, or SOX9 in the interior cells. Magnification 20.

The findings with respect to all colony types sug- The gallbladder does not contain peribiliary glands gests that colony centers contained more primitive cells (Figs. 1, S4) but does have related cells with weaker and those at the edges were slightly more differenti- levels of stem/progenitor markers, strong evidence of ated. Cells transferred to differentiation conditions proliferative capacity (e.g., high expression of Ki67, showed loss of EpCAM and acquisition of mature Fig. S4) but less able to give rise to adult cell types markers. With the colony type 3 cells the EpCAM was other than those of the biliary tree (data not shown). lost at the edges; acquired by cells interiorly; and Both RT-PCR and immunostaining data suggest that finally, with full differentiation, loss of EpCAM alto- they might be transit amplifying cells, a hypothesis gether. Thus, EpCAM appears to be an intermediate being tested. marker of differentiation. Transcription Factors Are Located Both Intranu- Cells in all three colony types were consistently neg- clearly and Perinuclearly. The various transcription ative by immunohistochemistry for mesenchymal factors (e.g., SOX17, SOX9, PDX1) were found pre- markers such as desmin, a-smooth muscle actin dominantly intranuclearly both in situ and in cultured (ASMA), markers of endothelia (e.g., CD31 and vas- cells (Figs. 2-4, S7). Within each peribiliary gland cular endothelial cell growth factor receptor [VEGFr]), there was heterogeneous expression of transcription and hemopoietic markers (e.g., CD45, CD34) (data factors and of cytoplasmic and membrane-associated not shown). stem cell markers, with some cells positive and others 2164 CARDINALE ET AL. HEPATOLOGY, December 2011

Fig. 4. In situ double immunofluorescence staining for SOX17 and PDX1 indicating coexpression in some cells (images represent PBGs within hepato-pancreatic ampulla). The percentage of SOX17þ and PDX1þ cells within PBGs is 10%-15% with SOX17þ cells constituting 11.2% 6 3.76% (standard deviation [SD]) and PDX1þ cells constituting the 16.6% 6 3.43% (SD). The heterogeneity in cellular subpopulations within the PBGs was not due to an artifact of sectioning, because the sections used were 3-5 lm thick and the cell diameters, in vitro and in vivo,is 7-9 lm. The analysis and the images were made with a wide field fluorescence microscope and not with a confocal microscope. Therefore, most of the nucleus is displayed in a single section with the signal representing the amount of antigen present in most of the nucleus. negative. Although most of these markers were shared multiple transcription factors relevant to liver and pan- by cell populations from all biliary tree sites examined, creas (e.g., HNF6, NGN3) in some peribiliary glands there were distinctions in the relative expression of one is a unique feature that is distinctive from findings versus another marker and in whether key transcrip- with respect to embryonic stem (ES) cells lineage tion factors (e.g., SOX 17, PDX1) were located within restricted to liver or pancreas and in which specific the nucleus (Figs. 2, 4) or perinuclearly (Figs. 3, S7). genes turn on (and then off) in stages. Marker analyses A perinuclear localization occurred in some cells completed to date of biliary tree stem/progenitors indi- in situ and in cells at the edges of the type 3 colonies. cate they are stages between definitive endoderm and The transition from intranuclear to perinuclear loca- determined stem cells and mostly at lineage stage 4 in tion is interpreted as sequestration and/or turnover of the development of the endocrine pancreas or of the transcription factors accompanying differentiation liver from ES cells.18 events. Alternatively, the perinuclear localization could Expansion Potential of the Biliary Tree Stem/ indicate that the factors are in an inactive storage form Progenitor Cells. Cultures of the biliary tree tissue that can be activated by translocation to the nucleus on plastic and in serum-free KM resulted in selection under appropriate regenerative demands. for colonies of cells that divided initially every 36- Interestingly, there were also cells coexpressing mul- 40 hours, thereafter slowing to a division every 2-3 tiple transcription factors such as SOX17 and PDX1 days, with proliferation continuing for months and (Fig. 4). The percentage of SOX17þ cells is 11.2% 6 associated with stable maintenance of the undifferenti- 3.8% and the PDX1þ cell is the 16.6% 6 3.4% and ated cell phenotype (Table S2). Figure S8 shows a rep- the percentage of the cells coexpressing SOX17 and resentative colony maintained for more than 8 weeks PDX1 was variable but ranged from 10%-15%. This on culture plastic and in KM. Cells in the colony cen- coexpression in some cells and, similarly, expression of ters (regions a and b) had an average cell diameter of HEPATOLOGY, Vol. 54, No. 6, 2011 CARDINALE ET AL. 2165

6-7 lm, whereas those at the colony edges were (Figs. 5, 6, S10) accompanied by significant increases larger (11-12 lm) (regions c-e). We sampled suffi- in hepatocyte-specific gene expressions that included cient regions to enable us to have an estimate of early (e.g., HNFa4, AFP, CK8 and 18, and albumin), 400,000-500,000 cells for the entire colony by intermediate or zone 2 (e.g., transferrin, tyrosine ami- Metamorph analyses; this was an underestimate notransferase [TAT]), and late or zone 3 genes (e.g., given the dense aggregates of cells in some regions of P450 3A4) (Fig. 7). the colony. Similar analyses were done on other colo- Lineage Restriction to Bile Ducts Lined with nies at 4-8 weeks of culture (Table S3) and indicated Cholangiocytes. The presence of cells expressing that the cells went through divisions every 3 days markers of cholangiocytes (CK7, secretin receptor such that by 2 months they had gone through 18-20 [SR], and CFTR) occurred minimally in the SC con- divisions. trol conditions with an average of 3.2% 6 2.6% Biliary Tree Stem Cell/Progenitors Lineage Restrict positive cells found in each colony. In cells on plastic Either to Hepatocytes, Bile Ducts, or Pancreatic Islets and in HDM-C, clusters of cells coexpressing CK7, Depending on the Microenvironment. The biliary tree SR, and CFTR were observed concentrated at the pe- stem/progenitors were maintained for 4-8 weeks in an riphery of the colonies and their numbers increased undifferentiated state in culture on plastic and in KM to 49.2% 6 11.1% of the cells/colony (Fig. S9). resulting in 100% of the cells of colony types 1 and The colony centers remained primarily as undifferen- 2 and 20%-30% of those in colony type 3 being tiated cells (negative for SR and CFTR and positive positive for EpCAM. At the timepoint of transfer to for EpCAM; data not shown). Quantitative RT-PCR culture conditions other than KM and plastic, the cells of the cultures in the SC controls versus those in in all colony types contained cells strongly expressing HDM-C indicated that SR messenger RNA (mRNA) markers of stem cells (e.g., CXCR4, SOX9, SOX17, more than doubled (P < 0.05). The effect was even PDX1, CD133) and negligible levels of expression of more profound in the cells embedded into matrix genes indicative of mature cells (e.g., albumin, secretin and given HDM-C, resulting in ramifying and receptor, insulin). The potential adult fates of biliary branching ducts formation, lined by cells with a phe- tree stem/progenitors were realized by passaging equal notype of mature cholangiocytes (Fig. 8C). Quantita- numbers of them from cultures in KM into one of tive RT-PCR indicated significantly higher levels of three distinct differentiation conditions tailored either expression of cholangiocyte-specific genes for GGT, for liver, bile duct, or pancreatic islets. Each condition CFTR, and AE-2 (Fig. 7). was comprised of a serum-free HDM tailored for the Lineage Restriction to Functional Pancreatic adult tissue of interest: HDM-L (hepatocytes), HDM- Islets. Formation of islet-like structures increased sig- C (cholangiocytes), and HDM-P (pancreatic islets). nificantly both in cultures on plastic and in HDM-P For the 2D cultures the cells were plated onto culture and when embedded into matrix and given HDM-P. plastic and in just HDM; for the 3D cultures the The monolayers in HDM-P produced dense balls of specific HDM was used in combination with embed- aggregated cells budding from the edges of the colo- ding the cells into a mixture of extracellular matrix nies and containing cells expressing C-peptide, PDX1, components also tailored for the desired adult cell and insulin. Four to five such aggregate-structures type. Passaging the cells again onto plastic and in KM appeared on average in cells on plastic and in HDM- resulted in self-replication, conditions used as the stem P; secreted human C-peptide could be detected espe- cell (SC) controls. cially in cultures stimulated with high glucose levels Lineage Restriction Yielding Hepatocytes. Cells (Fig. 5). In cultures embedded into matrix and given with hepatocyte markers did not occur in the SC con- HDM-P, the islet-like clusters occurred at the edges of trol conditions. The numbers of cells coexpressing the hydrogels (pale blue structures) and were positive CK18 and albumin increased to 36.7% 6 10.4% in for dithizone staining, indicating cells with zinc 2D (monolayer) cultures in HDM-L (Fig. S9A-C), condensed in insulin granules in the cytoplasm and present mostly at the periphery of the colony, (Fig. 6). Immunohistochemistry indicated that these whereas colony centers consisted primarily of undiffer- neoislet-like structures were positive for PDX1, human entiated cells (negative for albumin and positive for C-peptide, and insulin as well as for glucagon and EpCAM; data not shown). somatostatin (data not shown). The quantitative In the HDM-L and embedded into matrix in 3D, RT-PCR assays (Fig. 7) of these cultures were the most cords of cuboidal-shaped cells with ultrastructural dramatic in the elevation of expression of genes and functional features of hepatocytes were observed specific for pancreatic endocrine cells. 2166 CARDINALE ET AL. HEPATOLOGY, December 2011

Fig. 5. Differentiation toward a pancreatic islet fate in cells on plastic and in HDM-P. When the biliary tree stem cells were cultured as mono- layers in HDM-P the cultures changed within a week to have cell aggregation and condensation at the edges of the colonies, resulting in the for- mation of islet-like structures containing C-peptide (B,C) insulin (D), and PDX-1 (not shown). (A) Undifferentiated cells (EpCAMþ cells) were found within the colony centers. Magnification 20. The number of islet-like structures/colony increased from 1 6 0.7 to 3.8 6 1.3 in the cul- tures in HDM-P. (E) The levels of human C-peptide in ng/lg of at low (basal) glucose concentration (5.5 mM) for a 2-hour incubation in cultures in KM was 4.5 6 2.25 versus 12.3 6 1.9 ng/lg of protein in the cultures in HDM-P. The data are expressed as means 6 standard error; n ¼ 4; *P < 0.05. Glucose stimulated C-peptide secretion (F) was observed in the HDM-P with the levels of human C-peptide in the me- dium at 1.10 6 0.32 lg/L under low (basal) glucose concentration (5.5 mM) to 1.92 6 0.43 lg/L in high glucose concentration (22 mM); n ¼ 7; *P < 0.01.

Transplantation of Biliary Tree Stem/Progenitor were found around injection sites and some were dis- Cells In Vivo Results in Distinct Fates Depending on persed into the liver sinusoids of periportal and intra- Site and Microenvironment. To explore whether bili- lobular parenchyma. We estimate that an average of ary tree stem cells have the ability to differentiate into 6.52% 6 2.5% of the total area of the hepatic lobes mature liver cell type in vivo, we transplanted isolated was occupied by mature human hepatocytes positive biliary tree stem/progenitors directly into the livers of for human HepPar-1 and albumin (Fig. 8). Moreover, normal SCID mice (n ¼ 3) and checked for their the bile ducts were lined with a high percentage (aver- fate(s) 30 days later. The mice were not subjected to age 12.7% 6 5.5%) of mature human cholangiocytes, any injury process, meaning that the engraftment positive for human CK7 (Fig. 8), meaning that human occurred under quiescent liver conditions. Human cells cholangiocytes coexisted with human hepatocytes HEPATOLOGY, Vol. 54, No. 6, 2011 CARDINALE ET AL. 2167

Fig. 6. Adult fates of biliary tree stem cells when given a specific HDM and embedded in mixtures of extracellular matrix components. (A) Macroscopic images of branching and ramifying bile duct formation lined by layers of epithelial cells within 7-9 days in cultures of biliary tree stem cells under 3D conditions in MKM-C show the differentiation toward cholangiocytes. Magnifications are 10 and 20 for (A,B), respec- tively. (C,D) After 9 days of restriction under MKM-L conditions the differentiation toward hepatocytes, cords of cuboidal cells formed and were interspersed with bile canaliculi. Magnification 20 (see also Fig. S10 for ultrastructural images of the cells under the stem cell control condi- tion versus this 3D differentiation condition). (E-G) After 7-14 days of differentiation toward a pancreatic islet fate, islet-like structures are observed budding from the surface of the ball of cells and expressed PDX1 (pink) and C-peptide (blue/green). The nuclei are blue from DAPI staining. (E) Higher magnification images (20) show two of the islet-like structures that formed at the edges of the hydrogels at day 5 and day 7 in the 3D differentiation conditions. The aggregates became positive for dithizone staining by day 7, indicating expression of insulin. within areas of liver parenchyma in the hosts. Human maintained normal glucose levels throughout the cells were not observed in sham-transplanted animals, experiments. The sham controls given STZ became and no tumors formed in any of the transplanted hyperglycemic and within 2 weeks had glucose levels animals. at > 750 mg/dL. These controls maintained high lev- To assess ability to form functional pancreatic endo- els of hyperglycemia for the duration of the experi- crine cells in vivo, biliary tree stem/progenitors lineage ments and some of them died at around 100 days. By restricted partially toward a pancreatic islet fate were contrast, the glucose levels in STZ-treated mice and transplanted in Rag2 / /Il2rg / mice that were transplanted with preinduced neoislet clusters subsequently rendered diabetic by treatment with remained high (>750 mg/dL) for 2 months and streptozocin (STZ). The mice not subjected to STZ then declined steadily. By day 102 the glucose levels 2168 CARDINALE ET AL. HEPATOLOGY, December 2011

Fig. 7. Proof of multipotentiality as assessed by gene expression. Quantitative RT-PCR assays were done on biliary tree stem/progenitors main- tained under self-replication conditions (SC control) versus in one of three, distinct serum-free, hormonally defined medium (MKM-L, MKM-C, or MKM-P) and either on culture plastic or used in combination with embedding the cells into a specific combination of extracellular matrix compo- nents. The mRNA relative expression levels were calculated by normalization of DDCt values against the expression of glyceraldehyde 3-phos- phate dehydrogenase (GAPDH). The expression levels of a given gene under self-replication conditions are indicated as 1.0 and the levels under the other conditions are the fold difference. Below are summarized the conditions assayed; the number in parentheses is the number of times the experiment was done. 2D (monolayer) cultures on tissue culture plastic and in (A) KM (13); (B) HDM-L: albumin assays (9); (C) HDM-P: insulin assays (13); (D) HDM-C: secretin receptor (6). 3D cultures of cells embedded in mixtures of extracellular matrix components and in (E) HDM-C (3); (F) HDM-L (3); or (G) HDM-P (3). The genes assayed were GGT1: gamma glutamyl transpeptidase-1; AE2: anion exchanger 2; CFTR: cystic fibrosis transmembrane conductance regulator; HNF4a: hepatocyte nuclear factor 4A; AFP: alpha-fetoprotein; ALB: albumin; TF: transferrin; TAT: tyrosine aminotransferase; CYP3A4: cytochrome P450 3A4; : Pancreatic and duodenal 1; ISL-1: ISL LIM homeo- box 1; NGN3: neurogenin 3; INS: insulin; GCG: glucagon. were less than half that of the controls. All of these livers and canals of Hering in pediatric and adult liv- mice survived, and there was no tumor formation in ers.4,5,19,20 They start at the level of intrahepatic septal any of them. Significant levels of human C-peptide bile ducts, implicating these as additional intrahepatic were detected at postoperative days 68 and 91 in the stem cell niches, corroborating the findings of Theise serum of hosts transplanted but not control or sham et al.19 These multipotent stem cells, located in peri- control mice (P < 0.001). The human C-peptide biliary glands deep within the bile duct walls, express levels in vivo were regulatable by glucose challenge markers for endodermal stem cells and can migrate to (Fig. 8). appropriate sites and differentiate into various adult cells, contributing to the renewal/repair of biliary epi- Discussion thelium and also of liver and pancreas. Given that cells and the differentiation phenomena are found in biliary Peribiliary glands are stem cell niches of the biliary tree tissue from fetal, pediatric, adult, and geriatric tree and compare with and are related to intrahepatic donors, facets of organogenesis of liver, biliary tree, stem cell niches in ductal plates of fetal and neonatal and pancreas appear to be ongoing throughout life. HEPATOLOGY, Vol. 54, No. 6, 2011 CARDINALE ET AL. 2169

Fig. 8. Adult fates of the stem/progenitor cells as indicated by transplantation in vivo in immunocompromised mouse. (A,B) In the liver sec- tions prepared from the hosts, around 6.52% 6 2.5% of the total area was occupied by transplanted cells which were positive for human Hep- Par-1 (arrows). Magnification 20. (B) Double immunofluorescence revealed the colocalization of human HepPar-1 and human albumin. Magnification 20. (C) Immunohistochemistry of liver sections for human specific CK7 (arrows) indicative of cholangiocytes shows on average 12.7% 6 5.5% of all bile duct cells are derived from transplanted stem cells. It was found that 14.92% 6 5.9% cells in the large bile ducts and 5.02% 6 1.95% of the cells in small bile ducts were positive for human CK7. No immunoreactions were observed in controls (n ¼ 6). Magnification 40. (D) The epididymal fat pads (EFP) of male Balb/C Rag2//Il2rg/ mice were injected with 200-400 preinduced, biliary tree stem/progenitor derived cell aggregates in the 3D differentiation conditions. Glucose tolerance tests performed at postoperative days 68 and 91 showed significant blood levels of human C-peptide in experimental mice and these levels were regulatable by glucose.

The gallbladder does not contain peribiliary glands, progenitors, but its relevance is not yet fully but it does have related cells that possibly represent understood. facultative progenitors. This proposal parallels the Cultures of the biliary tree stem/progenitors were intestinal model in which proliferation of stem cells obtained readily in KM, a serum-free, defined medium within Lieberkuhn’s crypts is followed by cell migra- developed for rodent hepatoblasts and subsequently tion and differentiation along the crypt-villus axis found effective for hepatic stem cells.5 In these condi- and is critical for development of the intestinal tions the biliary tree stem/progenitors remained pheno- architecture.21 typically stable and underwent a division initially every SOX17 is important for endodermal progenitors 36-40 hours, but slowed to a division every 3 days by switching between biliary tree and pancreas,15 is asso- 4 weeks, and at all times assayed thereafter had a prolif- ciated with hedgehog known also as impor- erative rate comparable to that found for hHpSCs.5 tant for liver versus pancreas differentiation, and is Under all conditions and with all the types of stem/pro- associated with primary cilia.22 We assume this is - genitors identified, the edges of the colonies contained evant to the SOX17 evident in the biliary tree stem/ cells that were larger and more differentiated than those 2170 CARDINALE ET AL. HEPATOLOGY, December 2011 in the colony interiors mimicking the formation of liver matrix molecules; both the soluble and matrix signals tissue from ductal plates4 and that of islets from the were tailored for different adult cell types. The micro- edges of pancreatic ducts during organogenesis.11 environment drives the cells rapidly to a specific adult Biliary tree stem/progenitors are logical precursors fate that can be easily identified by morphology and for mature cells of liver, bile duct, and pancreas given functions. known events in organogenesis and in studies on path- Further differentiation occurs with or without ologies of these tissues.6 They overcome many, if not selective pressure in vivo. That for lineage restriction all, of the ongoing controversies about whether or not toward the intrahepatic parenchymal cell fates there are stem cells in adult tissues for pancreas.11 The resulted in rapid differentiation, because the in vivo inability to identify true stem cells in adult pan- environment is toward integration of the cells within creas10,11,23,24 has fueled attempts to lineage restrict the liver plates followed by differentiation and with ES cells, induced pluripotent stem (iPS) cells, amniotic minimal adverse variables. By contrast, that for pan- fluid-derived stem cells (AFSCs), or mesenchymal stem creatic islets in hosts with STZ-induced diabetes cells (MSCs), often with genetic manipulation to involved selective pressure for differentiation in com- direct these cells to an endodermal fate; to find ways bination with adverse effects of the toxic environment to reprogram adult pancreatic acinar cells; or to elicit induced by hyperglycemia. Blood vessels can become proliferation of existing pancreatic islet beta narrowed or blocked by a buildup of fat and choles- cells.11,25,26 The process of driving these various stem terol due to abnormal glucose metabolism and poor cell populations to a mature endodermal tissue fate is nutrient conditions in hyperglycemic environment. inefficient, requiring up to 4-6 weeks in culture, and Transplanted precursors for pancreatic islets can take yielding adult cells with muted functions or with over- months to fully mature given this mix of positive expression of or absence of expression of some genes. and negative effects resulting from hyperglycemia, as In the cases of the MSCs, the resulting adult cells are has been shown by others.28 Thus, the speed of dif- phenotypic hybrids of mesenchymal cells and the ferentiation achieved in vitro versus in vivo is distinct desired adult cell type.27 Moreover, for reasons for each adult fate being generated and dependent on unknown, the phenotype of the adult cells generated is the balance of positive and adverse effects of the in vivo distinct with every preparation and source. Fully microenvironment. mature hepatocytes or glucose-regulated b-cells from We interpret the survival and rescue of the trans- these precursors occur only with in vivo maturation of planted mice from hyperglycemia as due to the pro- transplanted cells after several months. Moreover, any duction of human insulin. One cannot exclude the therapy based on ES or iPS cells may be limited by possibility that some degree of spontaneous regenera- the potential of teratoma formation, usually ascribed tion of the mouse pancreas might have occurred, to residual undifferentiated stem cells.28,29 resulting in the regeneration of endogenous islets Our findings that there are endodermal stem/pro- releasing murine insulin, but even if this occurred it genitors in the biliary trees of all donor ages; that they was insufficient to rescue the sham controls. The clonogenically expand in vitro under wholly defined controls subjected to STZ maintained glucose levels conditions; and that they readily and efficiently lineage above 750 mg/dL throughout the experiments, and restrict to liver, biliary tree, or pancreatic adult fates in some of them died. One cannot consider the option culture with sets of wholly definable microenviron- of formation of hybrids between host pancreatic islet mental cues or in vivo suggest that they will become cells and the biliary tree-derived islet-like clusters, preferred choices for clinical programs and most because transplantation was done into the fat pads. experimental studies. The speed at which the differen- The strongest evidence is that the dramatic reduction tiation occurs, even in culture, is an indication that in hyperglycemia in the experimental mouse corre- these biliary tree stem/progenitors have progressed lated with increasing levels of human insulin/C-pep- through most of the developmental stages for a liver tide in their blood, and that the levels were regulat- or pancreatic fate. Indeed, their phenotypic character- able by glucose injected into the mice. The extent of istics in situ and in vitro indicate that they are mostly response observed seems logical given that low num- at stage 4 of the 5 stages during the progress of bers of partially differentiated neoislet-like clusters human ES cell stepwise-differentiation toward an islet were implanted; given the poor nutrient environment fate.18 The other factor in the speed of the response is for the grafts due to the high blood glucose levels; that we utilized a 3D microenvironment comprised of and given that mouse cells are far less sensitive to both a serum-free HDM and a specific mixture of human insulin than mouse insulin. We assume that a HEPATOLOGY, Vol. 54, No. 6, 2011 CARDINALE ET AL. 2171 normoglycemic state would be achieved within the 6. Turner R, Lozoya O, Wang YF, Cardinale V, Gaudio E, Alpini G, time frame we used if higher numbers of neoislets et al. Hepatic stem cells and maturational liver lineage biology. HEPATO- LOGY 2011;53:1035-1045. were transplanted. 7. Kubota H, Reid LM. Clonogenic hepatoblasts, common precursors for Our observations have pathophysiologic implications hepatocytic and biliary lineages, are lacking classical major histocompat- given that peribiliary glands, as stem cell niches, might ibility complex class I antigen. Proc Natl Acad Sci USA 2000;97: 12132-12137. play a role in injury repair as sites of origin of inflam- 8. Wang Y, Yao H-l, Barbier C, et al. Lineage-dependent epithelial-mesen- matory (primitive sclerosing cholangitis) or neoplastic chymal paracrine signals dictate growth versus differentiation of human (cholangiocarcinoma) diseases, paralleling the findings hepatic stem cells to adult fates. HEPATOLOGY 2010;52:1443-1454. 30 9. Wang Y, Cui C, Yamauchi M, Miguez P, Roach M, Malavarca R, et al. of Thayer and associates for pancreatic duct glands Lineage restriction of hepatic stem cells to mature fates is made 11 and the extensive studies by Bonner-Weir et al. efficient by tissue-specific biomatrix scaffolds. HEPATOLOGY 2011;53: on candidate progenitors in pancreatic ducts. It is 293-305. of interest that peribiliary glands have the highest 10. Dor Y, Brown J, Martinez O, Melton DA. Adult pancreatic beta-cells are formed by self-duplication rather than stem-cell differentiation. density at the hepato-pancreatic ampulla and common Nature 2004;429:41-46. hepatic duct at the hilum, sites at which cholangiocar- 11. Bonner-Weir S, Tosch E, Inada A, Reitz P, Fonseca SY, Aye T, http:// cinomas typically occur.31 The common embryologic www.ncbi.nlm.nih.gov/pubmed?term¼%22Sharma%20A%22%5BAuthor% 5D et al. The pancreatic ductal epithelium serves as a potential pool of pro- origin of intestine and biliary tree opens new pers- genitor cells. Pediatr Diabetes 2004;5:16-22. pectives on certain pathologies such as ulcerative colitis 12. Kuver R, Savard CE, Lee SK, Haigh WG, Lee SP. Murine gallbladder and sclerosing cholangitis or in the similarities between epithelial cells can differentiate into hepatocyte-like cells in vitro.AmJ 32 Physiol Gastrointest Liver Physiol 2007;293:G944-955. colorectal adenocarcinoma and cholangiocarcinoma. 13. Aikawa M, Miyazawa M, Okada K, Toshimitsu Y, Torii T, Otani Y, Comparisons of progenitor stem/populations in bili- et al. Regeneration of extrahepatic bile duct—possibility to clinical ary tree versus pancreas could provide explanations for application by recognition of the regenerative process. J Smooth Muscle the known distinctions in regenerative capacity of liver Res 2007;43:211-218. 14. Dutton JR, Chillingworth NL, Eberhard D, Brannon CR, Hornsey versus pancreas and could reveal if, as we suspect, MA, Tosh D, http://www.ncbi.nlm.nih.gov/pubmed?term¼%22Slack% organogenesis of liver and pancreas is ongoing 20JM%22%5BAuthor%5D et al. Beta cells occur naturally in extrahe- throughout life. These speculations are to be addressed patic bile ducts of mice. J Cell Sci 2007;120:239-245. 15. Spence JR, Lange AW, Suh-Chin J, Kaestner KH, Lowy AM, Kim I, with future studies. Biliary tree tissue is available from http://www.ncbi.nlm.nih.gov/pubmed?term¼%22Whitsett%20JA%22% fetal, neonatal, pediatric, and adult organs, including 5BAuthor%5D et al. SOX 17 regulates lineage segregation of ventral surgical materials (e.g., from cholecystectomy), tissue progenitor cells. Dev Cell 2009;17:62-74. 16. Furuyama K, Kawaguchi Y, Akiyama H, Horiguchi M, Kodama S, Kuhara routinely discarded from donor livers (gallbladder, T, http://www.ncbi.nlm.nih.gov/pubmed?term¼%22Hosokawa%20S% cystic ducts, periampular region), or pancreata (peri- 22%5BAuthor%5D et al. Continuous cell supply from a Sox 9-expressing ampular region, bile duct) rejected for use in trans- progenitor zone in adult liver, exocrine pancreas and intestine. Nat Genet plantation and made available for research. Thus, the 2011;43:34-41. 17. Kubota H, Reid LM. Clonogenic hepatoblasts, common precursors for extrahepatic biliary tree is an ideal and available source hepatocytic and biliary lineages, are lacking classical major histocompat- of stem/progenitor cells useful for regenerative medi- ibility complex class I antigens. Proc Natl Acad Sci U S A 2000;97: cine programs for liver, bile duct, and pancreas, 12132-12137. including for treatment of diabetes. 18. Murry CE, Keller G. Differentiation of embryonic stem cells to clini- cally relevant populations: lessons from embryonic development. Cell Mol Life Sci 2008;132:661-680. References 19. Theise ND, Saxena R, Portmann BC, Thung SN, Yee H, Chiriboga L, et al. The canals of Hering and hepatic stem cells in humans. HEPATO- 1. Crawford JM. Development of the intrahepatic biliary tree. Semin LOGY 1999;30:1425-1433. Liver Dis 2002;22:213-226. 20. Kuwahara R, Kofman AV, Landis CS, Swenson ES, Barendswaard E, 2. Roskams TA, Theise ND, Balabaud C, Bhagat G, Bhathal PS, Bioulac- Theise ND. The hepatic stem cell niche: identification by label retain- Sage P, et al. Nomenclature of the finer branches of the biliary tree: ing cell assay. HEPATOLOGY 2008;47:1994-2002. canals, ductules, and ductular reactions in human livers. HEPATOLOGY 21. Zbar AP, Simopoulos C, Karayiannakis AJ. Cadherins: an integral role 2004;39:1739-1745. in inflammatory bowel disease and mucosal restitution. J Gastroenterol 3. Nakanuma Y, Katayanagi K, Terada T, Saito K. Intrahepatic peribiliary 2004;39:413-421. glands of humans. I. Anatomy, development and presumed functions. 22. Wong SY, Reiter JF. The primary cilium at the crossroads of mamma- A review. J Gastroenterol Hepatol 1994;9:75-79. lian Hedgehog signaling. Curr Top Dev Biol 2008;85:225-260. 4. Zhang L, Theise N, Chua M, Reid LM. Human hepatic stem cells and 23. Stanger BZ, Tanaka AJ, Melton DA. Organ size is limited by the num- hepatoblasts: symmetry between liver development and liver regenera- ber of embryonic progenitor cells in the pancreas but not the liver. Na- tion. HEPATOLOGY 2008;48:1598-1607. ture 2007;445:886-891. 5. Schmelzer E, Zhang L, Bruce A, Wauthier E, Ludlow J, Yao HL, 24. Zhou Q, Brown J, Kanaredk A, Rajagopal J, Melton DA. In vivo http://www.ncbi.nlm.nih.gov/pubmed?- reprogramming of adult pancreatic exocrine cells to beta cells. Nature term¼%22Moss%20N%22%5BAuthor%5D et al. Human hepatic 2008;455:627-632. stem cells from fetal and postnatal donors. J Exp Med 2007;204: 25. Bonner-Weir S, Weir GC. New sources of pancreatic beta cells. Nat 1973-1987. Biotechnol 2005;23:857-861. 2172 CARDINALE ET AL. HEPATOLOGY, December 2011

26. Zhou Q, Law AC, Rajagopal J, Anderson WJ, Gray PA, Melton DA. 30. Strobel O, Rosow DE, Rahaklin EY, Lauwers GY, Trainor AG, Alsina A multipotent progenitor domain guides pancreatic organogenesis. Dev J, et al. Pancreatic duct glands are distinct ductal compartments that Cell 2007;13:103-114. react to chronic injury and mediate Shh-induced metaplasia. Gastroen- 27. Snykers S, De Kock J, Rogiers V, Vanhaecke T. In vitro differentiation terology 2010;138:1166-1177. of embryonic and adult stem cells into hepatocytes: state of the art. 31. Cardinale V, Semeraro R, Torrice A, Gatto M, Napoli C, Bragazzi Stem Cells 2009;27:577-605. MC, http://www.ncbi.nlm.nih.gov/pubmed?term¼%22Gentile%20R% 28. D’Amour KA, Bang AG, Eliazer S, Kelly OG, Agulnick AD, Smart NG, 22%5BAuthor%5D et al. Intra-hepatic and extra-hepatic cholangiocar- et al. Production of pancreatic hormone-expressing endocrine cells from cinoma: new insight into epidemiology and risk factors (review). World human embryonic stem cells. Nat Biotechnol 2006;24:1392-1401. J Gastrointest Oncol 2010;2:407-416. 29. Kroon E, Martinson LA, Kadoya K, Bang AG, Kelly OG, Eliazer S, 32. Chiu CT, Chiang JM, Yeh TS, Tseng JH, Chen TC, Jan YY, et al. et al. Pancreatic endoderm derived from human embryonic stem cells Clinicopathological analysis of colorectal cancer liver metastasis and in- generates glucose-responsive insulin-secreting cells in vivo. Nat Biotech- trahepatic cholangiocarcinoma: are they just apples and oranges? Dig nol 2008;26:443-452. Liver Dis 2008;40:749-754.