Articles
Directed differentiation of cholangiocytes from human pluripotent stem cells
© 2015 Nature America, Inc. All rights reserved.
Mina Ogawa1,2,11, Shinichiro Ogawa1,11, Christine E Bear3, Saumel Ahmadi3, Stephanie Chin3, Bin Li4, Markus Grompe4, Gordon Keller1,5,6,12, Binita M Kamath7,8,12 & Anand Ghanekar2,9,10,12 Although bile duct disorders are well-recognized causes of liver disease, the molecular and cellular events leading to biliary dysfunction are poorly understood. To enable modeling and drug discovery for biliary disease, we describe a protocol that achieves efficient differentiation of biliary epithelial cells (cholangiocytes) from human pluripotent stem cells (hPSCs) through delivery of developmentally relevant cues, including NOTCH signaling. Using three-dimensional culture, the protocol yields cystic and/or ductal structures that express mature biliary markers, including apical sodium-dependent bile acid transporter, secretin receptor, cilia and cystic fibrosis transmembrane conductance regulator (CFTR). We demonstrate that hPSC-derived cholangiocytes possess epithelial functions, including rhodamine efflux and CFTR-mediated fluid secretion. Furthermore, we show that functionally impaired hPSC-derived cholangiocytes from cystic fibrosis patients are rescued by CFTR correctors. These findings demonstrate that mature cholangiocytes can be differentiated from hPSCs and used for studies of biliary development and disease. Protocols to generate tissue-specific cell types from human embryonic stem cells (hESCs) and induced pluripotent stem cells (iPSCs) have facilitated the development of novel in vitro models of human development and disease that have opened new avenues for drug discovery and cell therapy. Of the different lineages that can be studied with this approach, those of the liver are of special interest as hepatocytes and cholangiocytes, which make up the liver’s biliary system, are primary targets of the adverse effects of drugs and inherited and infectious diseases1. Biliary disorders such as cystic fibrosis (CF) and Alagille syndrome are common causes of liver disease that often necessitate liver transplantation2. More complex biliary diseases, such as primary sclerosing cholangitis and biliary atresia, lack appropriate models for understanding their pathophysiology and for screening pharmacological agents. Most studies to date have focused on generating hepatocytes from hPSCs3–6 and have yielded protocols that give rise to relatively mature cells suitable for predictive drug toxicology and modeling certain liver diseases in vitro5,6. Recently, cells with cholangiocyte characteristics were generated from hPSCs7, demonstrating that it is possible to derive this lineage in vitro as well. However, because these populations were established as monolayers, their functional capacity remains unclear, as cells cultured in this format often do not accurately recapitulate the normal physiological responses of complex tissues. The generation of functional three-dimensional (3D) biliary structures from hPSCs by following developmentally relevant signaling cues would overcome this limitation and provide an in vitro model
for studying biliary diseases and a foundation for developing tissue replacement strategies. Cholangiocytes derive from a bipotential progenitor known as the hepatoblast that also gives rise to hepatocytes8. Targeting studies in mice have shown that cholangiocyte specification from the hepatoblast is mediated by the interaction of Notch2, expressed by the progenitors, and Jagged1, present on the developing portal mesenchyme9,10. Notch signaling also plays a critical role in intrahepatic bile duct morphogenesis in mice11,12. The discovery that the pediatric biliary disorder Alagille syndrome, characterized by bile duct paucity, is caused by mutations in JAG1 or NOTCH2 provides strong evidence that NOTCH signaling is also involved in cholangiocyte development in humans13–15. However, the role of the NOTCH pathway in regulating human cholangiocyte development from hPSCs has not been investigated. Here we developed a serum-free protocol that directs the efficient differentiation and maturation of functional cholangiocytes from hPSCs by delivering developmentally relevant cues, including NOTCH signaling through co-culture with OP9 stromal cells, at the hepatoblast stage of development. We show that hPSC-derived cholangiocytes grown in 3D culture form epithelialized cystic and/or ductal structures that express markers found in mature bile ducts, including the CFTR. Cholangiocytes from normal hPSCs generate cystic and/or ductal structures and display epithelial functions, including CFTR-mediated fluid secretion, observed through the regulation of cyst swelling, after stimulation of the cAMP pathway with forskolin. Cysts generated
1McEwen
Centre for Regenerative Medicine, University Health Network, Toronto, Ontario, Canada. 2Toronto General Research Institute, University Health Network, Toronto, Ontario, Canada. 3Program in Molecular Structure & Function, The Hospital for Sick Children Research Institute, Toronto, Ontario, Canada. 4Department of Pediatrics, Oregon Health and Science University, Portland, Oregon, USA. 5Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada. 6Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada. 7Division of Gastroenterology, Hepatology and Nutrition, The Hospital for Sick Children, Toronto, Ontario, Canada. 8Department of Pediatrics, University of Toronto, Toronto, Ontario, Canada. 9Division of General Surgery, University Health Network, Toronto, Ontario, Canada. 10Department of Surgery, University of Toronto, Toronto, Ontario, Canada. 11These authors contributed equally to this work. 12These authors jointly directed this work. Correspondence should be addressed to A.G. ([email protected]) or S.O. ([email protected]). Received 2 February; accepted 19 June; published online 13 July 2015; doi:10.1038/nbt.3294
nature biotechnology advance online publication
Articles
RESULTS Characterization of hPCS-derived hepatoblasts To generate hepatoblasts from hPSCs, we used our previously published protocol6 (Fig. 1a) with the modification that the endoderm induction step was carried out in monolayers rather than in three-dimensional (3D) embryoid bodies. With this approach, populations with >90% of cells expressing the definitive endoderm markers CXCR4, cKIT and EPCAM (ref. 16) were generated by day 7 of differentiation (Fig. 1b). Hepatic specification of this endoderm population was achieved by the addition of bFGF and BMP4 (refs. 6,16,17). Formation of the mouse embryonic liver bud is dependent on transient expression of Tbx3 (refs. 18,19). As the liver bud expands, progenitor cells downregulate Tbx3 and maintain and/or upregulate the expression of genes normally expressed in the hepatic and/or cholangiocyte lineages, including albumin (Alb), alpha fetoprotein (Afp)
Day 3 Day 5 Day 7
104 103 0.123
0102 103 104 105
cKIT 105
0.188
99.6
4
10
103 102 0
0.047
0.212
0
0
8 6
AFP
4 2 0
e
AFP/TBX3
f % of max.
100 80
% of max.
1.0
2
20
1
0.5
0
0
0
ALB
CK19
Merge
AFP 61.6
100 60
11.7
40 20
96
40 20 0
0 102 103 104 105 ALB
CK19
80
100
19.4
40 20 0
0 102 103 104 105
NOTCH2
0 102 103 104 105
AFP
100
80 60
96.6
40 20 0
0 102 103 104 105
80 60
100
100 80 60
CK19
80 60 40
94.1
20 0
0 102 103 104 105
0 102 103 104 105
ALB
100 90.4
40 20 0
**
1.5 NOTCH1
DAPI
Day 13
***
Merge
80
0
NOTCH2
AFP
20
60
3
ALB
40
100
4
*
DAPI
60
0
* *
5
40
**
2.5 ALB 2.0 1.5 1.0 0.5 0
***
0
80 CK19 60
***
ES
EPCAM
d
5
10 HNF6
*** *** ***
10
10
C D s a D y7 ay D 13 ay D 19 ay 25 AL FL
0 102 103 104 105
20 SOX9 15
***
% of max.
0.617
20
% of max.
102 0
30 TBX3
HGF/EGF/ TGFβ +/– OP9
% of max.
97.3
Fold difference relative to fetal liver
CXCR4
1.95
HGF/Dex/OSM
bFGF/BMP4
0 102 103 104 105
% of max.
c
105
Day 19
% of max.
b
Day 13
ES C D s a D y7 ay D 13 ay D 19 ay 25 AL FL
Activin CHIR
NOTCH signaling promotes cholangiocyte development To investigate the effect of NOTCH signaling on cholangiocyte development, we co-cultured aggregates generated from day-25 hepatoblast populations with OP9 stromal cells, which are known to express different NOTCH ligands including JAGGED1 Cholangiocyte specification (refs. 21–23). As shown by immunostaining Day 25 and flow cytometry, the majority of cells within
Hepatoblast differentiation
% of max.
Day 0
Hepatic specification
cytokeratin 19 (Ck19; also known as Krt19), SRY-BOX 9 (Sox9), hepatic nuclear factor 6 (Hnf6; also known as Onecut1) and Notch2 (refs. 8,20). Analyses of bFGF- and BMP4-treated hESC-derived endoderm revealed transient upregulation of TBX3 at day 13, identifying this time as the stage of hepatoblast specification (Fig. 1c). The majority of cells at day 13 expressed TBX3, indicating that hepatoblast specification was efficient (Fig. 1d). The onset of SOX9 and HNF6 expression overlapped with that of TBX3 (Fig. 1c). However, unlike that of TBX3, expression of these genes continued to increase until day 25, the final day of analysis. Expression of ALB and AFP was upregulated at day 19 and also increased at day 25. CK19 showed a biphasic pattern, with peak levels detected at days 13 and 25. Immunofluorescent staining (Fig. 1e) and flow cytometry (Fig. 1f) indicated that the majority of cells at day 25 were ALB+, AFP+ and CK19+. These findings strongly suggest that the day-25 cell populations represent the expanded hepatoblast stage of development, equivalent to the liver bud stage in vivo. NOTCH2 but not NOTCH1 expression was also upregulated at day 25, further demonstrating that this population contains hepatoblasts capable of signaling through this pathway.
ES C D s a D y7 ay D 13 ay D 19 ay 25 AL FL
Definitive endoderm monolayer
ES C D s D ay 7 ay D 13 ay D 19 ay 25 AL FL
a
CXCR4
© 2015 Nature America, Inc. All rights reserved.
from CF patient iPSCs show impairment in forskolin-induced swelling that is rescued by addition of clinically relevant CFTR modulators. These findings demonstrate that delivery of NOTCH signaling to hPSC-derived hepatoblasts drives the generation of structurally and functionally mature cholangiocytes that can be used to model clinically relevant biliary diseases in vitro.
NOTCH2
80 60
95.7
40 20 0
Day 25
0 102 103 104 105
Figure 1 Characterization of the hepatoblast stage of development in hPSC differentiation cultures. (a) Schematic representation of the differentiation protocol. (GSK3 inhibitor: CHIR99021), bFGF (basic fibroblast growth factor), BMP4, HGF (hepatocyte growth factor), Dex (dexamethasone), OSM (oncostatin M), EGF (epidermal growth factor), TGFβ (transforming growth factor beta), OP9 (OP9 stromal cells). (b) Flow cytometric analyses showing the proportion of CXCR4+, cKIT+ and EPCAM+ cells in day-7 populations. (c) RT-qPCR analyses of the expression of the indicated genes in populations at different stages generated from H9 hESCs as indicated in Figure 1a. Values are determined relative to TATA-binding protein (TBP) and presented as fold change relative to the expression in fetal liver, which is set as 1. AL, adult liver; FL, fetal liver. ANOVA with Tukey’s HSD test compared with day-7 endoderm: *P < 0.05, **P < 0.01, ***P < 0.001. Error bars, s.d. ± the mean of three independent experiments. (d) Immunostaining analyses showing the proportion of AFP+ and TBX3+ cells in the differentiating population at day 13 of culture. TBX3, Alexa 488 (green); AFP Cy3 (red). Scale bar, 100 µm. (e) Immunostaining analysis showing the proportion of ALB+, AFP+ and CK19+ cells in the differentiating population at day 25 of culture. ALB, Alexa 488 (green); AFP and CK19, Cy3 (red); nuclei, DAPI (blue). Scale bars, 100 µm. (f) Intracellular flow cytometric analysis showing the proportion of AFP+, ALB+, CK19+ and NOTCH2+ cells in the populations at days 13 and 25 of culture. Blue, positive cells; red, isotype controls.
advance online publication nature biotechnology
Articles
DAPI
DAPI
b ALB
ALB
AFP
CK19
Merge
Merge
CTRL
EGF
c
e
EGF TGFβ1
HGF EGF
5
SOX9
CK19
Fold difference relative to OP9 (–)
a
3
HES5
HEY1
**
4
**
**
***
2 1 0
HGF
HES1
HGF EGF TGFβ1
OP9 OP9 OP9 OP9 OP9 OP9 OP9 OP9 OP9 (–) (+) (+) (–) (+) (+) (–) (+) (+) GSI GSI GSI (+) (+) (+)
Figure 2 NOTCH signaling promotes CFTR DAPI cholangiocyte differentiation from hPSC-derived hepatoblasts. (a) Confocal 2 ALB microscopic images of immunostained HGF EGF TGFβ1 day-27 aggregates showing co-expression TGFβ1 1 of ALB+, AFP+ and CK19+ cells. ALB, GSI Alexa 488 (green); AFP and CK19, 0 ASBT DAPI Cy3 (red); nuclei, DAPI (blue). Scale bars, 15 CK19 ALB DAPI CK19 50 µm. (b) Immunostaining analysis showing 10 the presence of ALB+ and CK19+ cells in hepatoblast populations co-cultured α-tubulin 5 CK19 for 9 d (total culture time 36 d) on OP9 stromal cells in the indicated conditions. DAPI CTRL: control with no additional cytokines. ALB, Alexa 488 (green); CK19 Cy3 0 (red); nuclei, DAPI (blue). Scale bars, 100 µm. (c) Immunostaining analysis + + + showing the presence of CK19 /SOX9 cells (upper panel), CFTR cells (middle panel) and ASBT+ cells (lower panel) in hepatoblast populations co-cultured for 9 d (total culture time 36 d) on OP9 stromal cells in HGF, EGF and TGFβ1. SOX9, Alexa 488 (green); CK19, CFTR and ASBT, Cy3 (red); nuclei, DAPI (blue). Scale bars, 100 µm. (d) Confocal microscopy images for immunostaining of acetylated α-tubulin localized in primary cilia on H9-derived cholangiocytes cultured on OP9 stroma in HGF, EGF and TGFβ1. Acetylated α-tubulin, Cy3 (red); nuclei, DAPI (blue). Inset image with triple staining confirms colocalization of CK19 and primary cilia. CK19, Alexa 488 (green). Scale bar, 10 µm. (e) RT-qPCR-based analysis of the expression of NOTCH target genes in hepatoblast populations after 9 d of culture (total 36 d) on OP9 stroma (OP9+) with or without GSI or Matrigel-covered wells (OP9 −) in the presence of HGF, EGF and TGFβ1. Values were determined relative to TBP and presented as fold-change relative to the levels in the cells cultured on Matrigel for 9 d (OP9 −), which is set as 1. Bars in all graphs represent the s.d. ± the mean of three independent experiments. Student’s t-test compared with the culture on OP9 stroma (OP9 +): **P < 0.01, ***P < 0.001 (n = 3). (f) RT-qPCR-based expression analysis of ALB and CK19 in hepatoblast populations cultured on OP9 stroma in the presence of HGF, EGF and TGFβ1 for 3 d (30 d total), 6 d (33 d total) and 9 d (36 d total). The addition of GSI is indicated. Hepatoblast (HB: day 27) represents the population before co-culture. Values were determined relative to TBP and presented as fold change relative to levels in the day 27 hepatoblast aggregates, which is set as 1. H Da B ( y d Da 30 ay 2 y G 7 Da 33 SI ) y G (– Da 36 SI ) y G (–) Da 30 SI y G (– Da 33 SI ) y G (+) 36 S G I(+) SI (+ ) AL FL
© 2015 Nature America, Inc. All rights reserved.
d
Fold change relative to hepatoblast
f
the aggregates before co-culture were ALB+AFP+CK19+NOTCH2+, indicating that they maintained hepatoblast characteristics (Fig. 2a and Supplementary Fig. 1a). When co-cultured on OP9 cells for 9 d, the aggregates formed distinct clusters of CK19+ cells that downregulated ALB expression, suggesting that they had undergone the initial stage of cholangiocyte specification (Fig. 2b, CTRL). As studies in mice have shown that Hgf, Egf and Tgfβ1 signaling plays a role in bile duct development8,20,24, we next added combinations of these factors to the cultures to determine whether these pathways would promote further development of CK19+ clusters. The addition of EGF alone, TGFβ1 alone or the combination of EGF, TGFβ1 and HGF led to a 3.0- to 3.5-fold increase in cell number compared with nontreated controls after 9 d of culture (Fig. 2b and Supplementary Fig. 1b). HGF alone had little effect. The majority (90%) of the cells in the populations cultured under these conditions were CK19+ and ALB−, indicating efficient specification and development of the cholangio cyte lineage (Supplementary Fig. 1c). The combination of either EGF and HGF or EGF, HGF and TGFβ1 induced substantial morphological changes and promoted the formation of CK19+ branched structures (Fig. 2b). CK19+ cells in structures treated with all three factors expressed other proteins indicative of the cholangiocyte lineage, including SOX9, the CFTR and the apical sodium-dependent bile acid transporter (ASBT) (Fig. 2c). They also displayed primary cilia as detected with acetylated α-tubulin (Fig. 2d). CFTR, ASBT and α-tubulin were not expressed in aggregates before OP9 cell co-culture (Supplementary Fig. 2). These findings demonstrate that co-culture of hepatoblasts with OP9 cells in the presence nature biotechnology advance online publication
of EGF, HGF and TGFβ1 promotes the development of a population that expresses cholangiocyte markers. With the efficiency of lineage specification in the co-culture model, we generated, on average, 4.16 ± 0.64 CK19+ALB− cholangiocyte-like cells per day-25 hepato blast and 6.42 ± 0.53 cholangiocyte-like cells per input hESC. Expression of NOTCH targets HES1, HES5 and HEY1 was upregulated after 9 d of culture on OP9 cells in the presence of EGF, HGF and TGFβ1 (Fig. 2e). This increase in expression was blocked by gamma secretase inhibitor (GSI), a NOTCH pathway antagonist, demonstrating that co-culture with OP9 cells effectively activated the NOTCH pathway. GSI also blocked downregulation of ALB, reduced the proportion of CK19+ cells and inhibited the development of branched structures in the cultures, indicating that these effects were mediated by NOTCH signaling (Fig. 2b and Supplementary Fig. 1c). RT-qPCR analyses (Fig. 2f) confirmed the immunostaining and flow cytometric findings and demonstrated a dramatic reduction in ALB and an upregulation of CK19 after co-culture with OP9 cells in the presence of EGF, HGF and TGFβ1. These expression patterns were reversed with GSI. Collectively, these findings indicate that activation of NOTCH signaling in the hepatoblast population induces the initial stages of cholangiocyte development and the combination of HGF, EGF and TGFβ1 signaling promotes the formation of branched structures reminiscent of early duct morphogenesis. Differentiation of two other hPSC lines (ESC HES2 and MSC-iPSC1) revealed similar temporal patterns of TBX3 and hepatoblast marker expression (Supplementary Fig. 3a). Aggregates of hepatoblasts from both generated branched structures consisting of CK19 +ALB− cells
Articles b Mix
Cyst
Tubular
Sphere
c 100
% of each type of structure
a
d
90 80 70 60 50 40 30 20 10 0
Mix Cyst Tubular Sphere
3D GSI(–)
3D GSI(+)
DAPI
CK19
ZO-1
Merge
DAPI
CK19
CFTR
Merge
© 2015 Nature America, Inc. All rights reserved.
GSI(–)
Figure 3 3D culture promotes cholangiocyte maturation. (a) Photomicrographs of duct-like and cyst-like structures that develop in the collagen/Matrigel cultures after 17 d of culture of chimeric aggregates consisting of H9-derived hepatoblasts and OP9 stromal cells (4:1 ratio) in the absence of GSI. Spheres DAPI SCTR CK19 Merge are dense colonies that develop in the presence of GSI. Scale bar, 100 µm. (b) Low-magnification photomicrograph showing cyst-like structures that develop from chimeric aggregates consisting of H9-derived hepatoblasts and OP9 stromal cells (4:1 ratio) after 17 d of culture in the collagen/Matrigel cultures. Scale bar, 100 µm. (c) Proportion of structures with tubular, cystic, DAPI ASBT CK19 Merge mixed and sphere morphology that developed per well from chimeric aggregates cultured in the absence or presence of GSI. Values represent averages from five DAPI ALB CK19 Merge independent experiments. (d) Confocal microscopy images of immunostained histological sections of 3D ductal and/or cyst structures after 17 d (total 44 d) GSI(+) of culture in the presence of HGF, EGF and TGFβ1 without GSI (GSI(−)). First and second rows: CK19, Alexa 488 (green); ZO-1 and CFTR, Cy3 (red). Third and fourth rows: SCTR and ASBT, Alexa 488 (green); CK19, Cy3 (red). Bottom ES OP9 OP9 row: confocal microscopic images of immunostained histological sections of 3D (kD) cells (+) (–) CK19 sphere structures after 17 d (total 44 d) of culture in the presence of HGF, EGF α-tubulin 250 and TGFβ1 with GSI (GSI(+)); ALB, Alexa 488 (green); CK19, Cy3 (red). In all DAPI B and C images, the nuclei are stained by DAPI (blue). Scale bars, 50 µm. (e) Confocal microscopy images of immunostaining of acetylated α-tubulin localized in α CFTR primary cilia present in a 3D ductal structure after 17 d (total 44 d) of co-culture 130 with OP9 stromal cells in the presence of HGF, EGF and TGFβ1. Acetylated α-tubulin, Alexa 488 (green); CK19, Cy3 (red); nuclei, DAPI (blue). Scale bar, 95 α CNX 10 µm. (f) Western blot analysis shows the presence of the mature glycosylated form of the CFTR protein (band C) in MSC-iPSC1-derived ductal and/or cyst structures generated after 17 d of culture of chimeric aggregates (OP9+) or aggregates of hepatoblasts without OP9 (OP9 −). Undifferentiated iPSCs were used as negative control and calnexin (CNX) was measured as a loading control.
e
following co-culture with OP9 stroma (Supplementary Fig. 3b), demonstrating the applicability of the approach to different hPSC lines. 3D culture promotes cholangiocyte maturation To determine whether growth in 3D structures would promote further maturation of the hPSC-derived cholangiocytes, we next generated chimeric aggregates consisting of day-25 hepatoblasts and OP9 (GFP +) stromal cells (Supplementary Fig. 4a) and cultured them in gels composed of 1.2 mg/ml collagen type-1, 40% Matrigel and containing HGF, EGF and TGFβ1. Within 2 weeks of culture, the aggregates gave rise to 3D structures that displayed either a tubular and ductal morphology, a hollow cyst morphology or a mixture of both (Fig. 3a–c). Expression of NOTCH target genes HES1, HES5 and HEY1 was significantly higher in structures derived from chimeric aggregates than in those aggregates without OP9 cells, indicating that NOTCH signaling was active under these conditions (Supplementary Fig. 4b). Approximately 17 (17.1 ± 6.06) structures developed per 105 day-25 input hepatoblasts, the majority (62.9 ± 30.7%) displaying the cyst morphology. The generation of these cyst and duct structures was dependent on NOTCH signaling, as they did not develop in the presence of GSI. Rather, under these conditions, we observed the formation of dense colonies that we refer to as spheres. Histological analyses revealed that tubular and cystic structures at day 44 had a lumen and contained epithelial-like cells that expressed
f
CK19 and E-CADHERIN but not ALB (Fig. 3d and Supplementary Fig. 4c). These cells also expressed the tight junction marker ZO-1 (Zonula occludens 1) CFTR, SCTR (secretin receptor) and ASBT and contained primary cilia (Fig. 3d,e). Most of these markers were restricted to the apical side of the structures, suggesting that the cells acquired apicobasal polarity, a feature of mature epithelial ducts. ASBT was also detected on the apical side of branching 3D structures (Supplementary Fig. 4d). The presence of only small numbers of Ki-67+ cells in CK19+ duct structures at this stage suggests limited proliferative capacity (Supplementary Fig. 4e). Similar patterns of CK19, CFTR and α-tubulin staining were observed in bile ducts of normal adult liver (Supplementary Fig. 5). Western blot analysis, RT-qPCR and flow cytometric analyses confirmed the presence of CFTR in cells generated from chimeric aggregates (Fig. 3f and Supplementary Fig. 6a,b). The levels of CFTR mRNA and protein were considerably higher in cells derived from chimeric aggregates than in those cultured without OP9 cells, indicating that CFTR expression was NOTCH-dependent. The spheres that developed in the presence of GSI expressed ALB and low levels of CK19, indicating that, in the absence of NOTCH signaling, cells with hepatoblast characteristics persist (Fig. 3d). RT-qPCR analyses showed similar changes as expression levels of hepatocyte markers (ALB, AFP and CYP3A7) increased whereas cholangiocyte-associated markers (CK19, SOX9 and CFTR) decreased (Supplementary Fig. 6c) after addition of GSI. advance online publication nature biotechnology
2.0 SOX9 1.5
1.5
*** ***
1.0
**
0.5
0.5 0 20
1.5
HNF1B
1.0
**
**
***
3
SCTR
2
10
** *
0 1.5 1.0
3
**
*
**
***
0 6
ASBT
*
**
4 2
0
15
*
5
1
0
1.5 TGR5
*
GGT
CFTR
20
2
0.5 0 20
0 25
10
0 4 SLC4A2/AE2
AQP1
1.0
15
1
5
HNF6
0.5
0 CK19
15
**
*
1.0
10 0.5
5
2.5
SSTR2
**
2.0 1.5
**
**
1.0 0.5
**
hE
C
H
hPSC-derived cholangiocytes form ductal structures in vivo To evaluate the developmental potential of hPSC-derived cholangiocytes in vivo, we transplanted 106 cells in a Matrigel plug into mammary fat pads of nonobese diabetic–severe combined immunodeficient–interleukin 2rγ−/− (NOD-SCID-IL2rγ−/−; NSG) mice. For these studies, we used dissociated cells from branched structures generated by co-culture of day-25 hepatoblasts derived nature biotechnology advance online publication
0 100
ALB
***
**
6,000 4,000
50
2,000 0
ep S ep ato i C at bla PS o h b s C C ola las t (H ho n t 9 la gio (iP ) n S Pr gio cyt C) c e i So ma yte (H rte ry ( 9) d hep iPS ch a C ol to ) an cy gi te oc yt e
To further validate the utility of the 3D 1 protocol, we next applied it to the human iPSC line MSC-iPSC1. A distribution of 0 CK19+/CFTR+/ASBT+ tubular and/or cystic structures similar to that observed in the H9 cultures developed from MSC-iPSC1-derived hepatoblast aggregates (Supplementary Fig. 7a). The efficiency of tubular and/or duct/cyst formation was also similar to that of the H9 hESCs, with 105 day-25 iPSC-derived hepatoblasts giving rise to 20.0 ± 1.64 structures. Primary cilia were also detected in these duct structures (Supplementary Fig. 7b). RT-qPCR analyses revealed that expression of genes encoding proteins found in functional cholangiocytes in vivo, including SOX9, HNF1B, HNF6, CK19, CFTR, SCTR, aquaporin-1 (AQP1), the Cl−/HCO3− anion exchanger 2 (SLC4A2/AE2), ASBT, gammaglutamyltransferase (GGT), TGR5 and the somatostatin receptor 2 (SSTR2) was significantly upregulated in hPSC-derived cystic structures compared to day-25 hepatoblasts (Fig. 4). Notably, the levels found in these structures were equivalent to or greater than those detected in primary cholangiocytes isolated by fluorescenceactivated cell sorting (FACS) from human liver using the ductal antibody DHIC5-4D9 (ref. 25; Supplementary Figs. 5 and 8). In contrast, HNF4A and ALB were expressed at higher levels in primary cultured hepatocytes and hPSC-derived hepatoblasts than in the cysts or primary cholangiocytes (Fig. 4 and Supplementary Table 3). Together, these findings show that when cultured in a mixture of Matrigel and collagen, the hepatoblast population can generate duct-like structures that express mature biliary markers.
***
***
0
hE H ep S ep ato i C P b a C to las SC h b C ola las t (H ho n t la gio (iP 9) n Pr gio cyt SC e ) i So ma cyte (H rte ry (iP 9) h d ep S ch a C ol to ) an cy gi te oc yt e
**
2
0 10,000 AFP 8,000
H
HNF4A
hE H ep S ep ato i C P b a C to las SC h b C ola las t (H ho n t la gio (iP 9) n Pr gio cyt SC e ) i So ma cyte (H rte ry (iP 9) h d ep S ch a C ol to ) an cy gi te oc yt e
0 3
H
Fold difference relative to sorted cholangiocyte
Figure 4 Gene expression profile of hPSCsderived cholangiocytes compared to sorted cholangiocytes. RT-qPCR analyses of the expression of the indicated genes in the indicated populations. hESC: undifferentiated H9 cells; iPSC: undifferentiated MSCiPSCs; hepatoblasts: day-25 populations; cholangiocytes: cystic structures generated from chimeric aggregates of either H9 or MSC-iPSCs-derived hepatoblasts; primary cultured hepatocytes; sorted cholangiocytes: cholangiocytes isolated from human liver samples by FACS, using the ductal-specific monoclonal antibody DHIC5-4D9 (ref. 25). Data are expressed as the fold change relative to sorted human cholangiocytes, which is set as 1. Student’s t-test: *P < 0.05, **P < 0.01, ***P < 0.001 (n = 3). Error bars in all graphs represent the s.d. ± the mean of three independent experiments. SCTR (secretin receptor), AQP1 (aquaporin 1), SLC4A2/AE2 (Anion exchange protein 2), ASBT, GGT (gamma-glutamyl transpeptidase), TGR5 (G protein- coupled bile acid receptor 1), SSTR2 (somatostatin receptor 2), HNF4A, AFP, (α-fetoprotein), ALB.
H
© 2015 Nature America, Inc. All rights reserved.
Articles
from HES2-RFP hESCs26 with OP9 stroma. Six to 8 weeks after transplantation, multiple duct-like structures were detected in the Matrigel plug (Supplementary Fig. 9a). Cells within the ducts were RFP+, demonstrating that they were of human origin (Supplementary Fig. 9b,c). Additionally they expressed CK19, CFTR and primary cilia, indicating that they displayed characteristics of cholangiocytes (Supplementary Fig. 9d,e). Teratomas were not observed in any of the animals receiving transplants. hPSC-derived cholangiocytes are functional To assess function of the hPSC-derived cholangiocytes, we evaluated their ability to efflux rhodamine 123, a dye used to measure activity of the multidrug resistance protein 1 (MDR1) transporter in normal bile duct cells. The cystic structures derived from H9 hESCs and normal iPSCs transported dye to the luminal space, indicative of transporter activity (Fig. 5a). In the presence of 20 µM verapamil, an inhibitor of MDR1, rhodamine did not accumulate in the lumen, confirming that dye movement reflected active transport through MDR1. To demonstrate CFTR functional activity, we performed a forskolininduced swelling assay on cystic structures (Fig. 5b). With this assay, activation of the cAMP pathway by addition of forskolin/IBMX increases CFTR function, resulting in fluid transport and cyst swelling27 that can be visualized after staining with the cellpermeable fluorescent dye calcein green. Addition of forskolin/IBMX to the cultures induced a fold increase of 1.23 ± 0.09 and 1.40 ± 0.17 in the size of H9- and iPSC-derived cysts, respectively, when measured 1 h later. Addition of the CFTR inhibitor CFTRinh-172 blocked the forskolin/IBMX-induced swelling, indicating that the increase
Articles b
MSC-iPSC1
c
(+)
MSC-iPSC1 500
***
300
*
***
***
***
400
***
300
200
200
100
100
0 FSK/IBMX (+) CFTR inhibitor (–)
MSC-iPSC1-derived cholangiocytes
Verapamil
H9 400
0 (+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
Figure 5 Function of hPSC-derived cholangiocytes in vitro. (a) Phase contrast images showing uptake of rhodamine 123 dye in 17-d-old cysts generated from cholangiocytes derived from either H9 hESCs or MSC-iPSC1 cells. To demonstrate functional transport, cysts were incubated with verapamil for 30 min before the addition of rhodamine 123. Images were taken 10 min after the addition of rhodamine (green). Scale bars, 100 µm. (b,c) Forskolininduced swelling assay demonstrating functional CFTR activity in hPSC-derived cysts. (b) Representative epifluorescence microscopy images of hESC and MSC-iPSC1-derived cysts before (−) and 24 h after (+) stimulation with forskolin/IBMX (F/I). Cells were labeled with calcein green and images were taken 24 h after F/I stimulation. Scale bars, 500 µm. (c) Quantification of the degree of cyst swelling 24 h after F/I stimulation in the absence or presence of the CFTR inhibitor CFTR Inh-172 (30 µM). Cyst swelling was quantified using velocity imaging software. Values shown are percent swelling relative to the cyst size before stimulation. Analyses from three independent experiments are shown. Student’s t-test: *P < 0.05, ***P < 0.001 (n = 3).
in cyst size was CFTR-dependent (Supplementary Fig. 10). When measured 24 h later, the swelling of H9- and iPSC-derived cysts increased 2.09 ± 0.21- and 2.65 ± 0.31-fold, respectively (Fig. 5c and Supplementary Movies 1 and 2). This swelling was also blocked by addition of CFTRinh-172. These findings demonstrate that cells in hPSC-derived cyst and/or duct-like structures display properties of functional cholangiocytes.
CF patients carrying the common F508del mutation (C1-iPSCs and GM00997-iPSCs). Both patient-derived iPSC lines generated definitive endoderm and hepatoblasts with kinetics similar to that of normal iPSCs (Fig. 6a,b and data not shown). Although hepatoblast development was not altered, cyst formation from the CF-iPSCs was clearly impaired as only branched structures were observed in gels after 17 d of culture (Fig. 6c). Cyst formation could be induced by addition of forskolin for the first week of the 2-week culture period (Fig. 6d). However, even under these conditions many cysts that developed from patient iPSCs were not completely hollow and contained branched ductal structures (Fig. 6e). As with the wild-type
10
3
98.8
2
10 0 0.2
0
2 0 10 103 104 105
cKIT 105 0
99.7
104 103 2
10 0 0
0.28
2 3 4 5 0 10 10 10 10
EPCAM
* ***
6 TBX3*** 4
*
2 0
25 AFP 20 15 10 5 0 5 SOX9 4 3 2 1 0
60
*** *
c
CK19
*
*
40
–FSK
20
*** ***
0 8
NOTCH2
6
***
4 2
ay
D
d
*** ***
D
ay
19
ay
D
19
MSC-iPSC1 Inh (+) Inh (–)
100
100
80
80
60
60
40
40
20
20
0
0
+FSK
0
19
C1-iPSC FSK (–) FSK (+)
7 d 14 d 7 d 14 d
4
ALB
Cyst formation (%)
10
2.5 2.0 1.5 1.0 0.5 0
C
CXCR4
105 1.04
Fold difference relative to fetal liver
b
1iP D SC D ay 7 ay 13 D ay 25 AL FL C 1iP D SC a D y7 ay 13 D ay 25 AL FL C 1iP D SC D ay ay 7 13 D ay 25 AL FL
a
e
MSC-iPSC1
7 d 14 d 7 d 14 d
Analyses of cholangiocytes generated from CF patient iPSCs To demonstrate the utility of this system to model disease in vitro, we analyzed cyst formation from iPSCs generated from two different
CXCR4
© 2015 Nature America, Inc. All rights reserved.
(–)
H9-derived cholangiocytes
DMSO
H9
% swelling relative to noninduced cysts
a
C1-iPSC
Figure 6 Generation of cholangiocytes from CF patient iPSCs. (a) Flow cytometric analyses showing the proportion of CXCR4+, cKIT+ and EPCAM+ cells in day-7 monolayer populations generated from a CF-iPSC line (C1 CFTR del). (b) RT-qPCR-based analyses of the expression of the indicated genes in populations at different stages generated from C1-iPSC. Cells were cultured as indicated in Figure 1a. Values shown are relative to TBP and presented as fold change relative to expression in fetal liver, which is set as one. AL: adult liver, FL: fetal liver. ANOVA with Tukey’s HSD test compared with day-7 endoderm: *P < 0.05, ***P < 0.001 (n = 3). Bars in all graphs represent the s.d. ± the mean of three independent experiments. (c) Photomicrographs of branched and cyst-like structures that develop in the collagen/Matrigel cultures after 17 d of culture of chimeric aggregates generated with cholangiocytes from CFTR F508del iPSCs (CF-iPSCs: C1-iPSC). Images are shown for structures that developed in the absence or presence of forskolin (FSK, 10 µM). Forskolin was added for the first week of the 17-d culture period (right). Scale bars, 100 µm. (d) Quantification of the number of cyst structures that developed from CF-iPSCs (C1-iPSC)-derived cholangiocytes cultured in the absence and presence of forskolin and from wild-type iPSCs (MSC-iPSC1) cultured in the absence and presence of the CFTR inhibitor CFTR Inh-172 (30 µM). Cyst structures were counted at 7 d and 14 d of culture. Forskolin was added to the CF-iPSC-derived aggregates and the CFTR inhibitor was added to the iPSC-derived aggregates for the first week of the 17-d-culture period (left graph). Bars represent the s.d. of the mean of three independent experiments. (e) Histological analyses showing the structure of iPSC- and CF-iPSC-derived cysts after 17 d of culture in the collagen/Matrigel gels (H&E staining; scale bars, 100 µm). Forskolin was included for the first week of the 17-d culture period.
advance online publication nature biotechnology
Articles Figure 7 Functional analysis of cholangiocytes from CF patient iPSCs. (a) Confocal images of histological sections showing the (–) Correction (+) Correction presence of the mutant CFTR protein in 16-d-old cyst structures (MW) C1-iPSC GM00997 – HBE generated from CF-iPSC-derived cholangiocytes cultured as chimeric (kD) iPSC aggregates in the presence or absence of the chemical correctors 250 α CFTR C VX-809 (3 µM) and Corr-4a (5 µM). The correctors were added at 2 weeks of culture. Forty-eight hours after the addition of the B (+) Correction 130 A correctors, the structures were fixed and analyzed for the presence α CNX of CFTR. Lower left panel shows the presence CFTR protein (Cy3, red) 95 (–) (+) (–) (+) in the cysts treated with the correctors. Membrane localized CFTR was VX-809 + Corr-4a not detected in the nontreated cysts (upper left panel). Right panel DAPI CK19 CFTR shows co-expression of CFTR and CK19 in a treated cyst. CK19, GM00997-iPSC Alexa 488 (green); CFTR, Cy3 (red); nuclei, DAPI (blue). Scale bars, (–) (+) 250 *** *** *** 50 µm. Arrows indicate a perinuclear distribution pattern of CFTR GM00997-iPSC 200 protein. (b) Western blot analysis showing the presence of the mature 150 glycosylated (band C) and core glycosylated (band B) form of the CFTR 100 protein in the nontreated and VX-809/Corr-4a-treated 16-d-old cysts. 50 HBE were used as a positive control. (c) Representative epifluorescence 0 microscopy images of calcein-green-labeled cyst structures generated C1-iPSC from cholangiocytes derived from CF-iPSC from two different patients * *** *** C1-iPSC (C1-iPSC and GM00997-iPSC). Images shown are before (−) and 24 h 400 after (+) stimulation with forskolin, IBMX and the CFTR potentiator 300 VX-770 (1 µM). Chimeric aggregates were generated in collagen/Matrigel 200 gels in the presence of forskolin for the first week of the 16-d culture period. 100 The correctors were added to the 2-week-old cultures for 48 h. At this stage, the cultures were stimulated with the combination of F/I and the FSK/IBMX 0 (+) (+) (+) (+) (+) (+) plus VX-770 potentiator VX-770. Cysts were measured 24 h later. Scale bars, 500 µm. VX-809 + Corr-4a (+) (–) (+) (–) (+) (–) (d) Quantification of the degree of swelling of the cysts derived from two different CF-iPSC lines. Cysts were cultured in the presence and absence of correctors, as indicated. Cyst swelling was quantified using velocity imaging software. The total size of the cyst was normalized to that before F/I stimulation from each of three individual experiments. Student’s t-test: *P < 0.05, *** P < 0.001 (n = 3).
b
c
d
© 2015 Nature America, Inc. All rights reserved.
% swelling relative to noninduced cysts
a
cells, CF iPSC-derived cysts contained primary cilia (Supplementary Fig. 11). A higher frequency of hollow cysts was detected following longer periods of culture, suggesting that maturation of the mutant cells was delayed. Addition of CFTR inhibitors to cultures of normal iPSC-derived cholangiocytes also delayed cyst formation, indicating that generation of these structures was dependent, to some degree, on functional CFTR at the cell surface (Fig. 6d). We assessed functional restoration of F508del CFTR in cholangio cytes from CF-patient iPSCs by adding the chemical correctors VX-809 and Corr-4a to cultures 2 d before the CFTR functional assay27–30. Both molecules correct folding defects of mutant CFTR protein. The addition of these correctors did not improve cyst formation but did result in accumulation of detectable levels of CFTR on the apical side of the lumen. Unlike the homogeneous distribution of CFTR in wild-type cysts, the mutant protein localized to the plasma membrane in some areas and exhibited perinuclear distribution in other regions (Fig. 7a). This pattern may reflect incomplete rescue of the trafficking defect of the mutant protein. Western blot analyses also revealed the effect of correctors. Wild-type CFTR normally migrates as two bands, one ~130–150 kD (band B) and the other ~170–190 kD (band C). This typical CFTR protein migration pattern is shown for lysates from human bronchial epithelial (HBE) cells (Fig. 7b). Band C represents the mature, Golgi-modified CFTR and band B, the immature, ER-localized CFTR protein. The F508del-CFTR protein migrated as a 130-150 kD protein in the iPSC-derived cholangiocytes in the absence of correctors consistent with retention in the endoplasmic reticulum (ER) and failure to traffic to the cell surface. Treatment with correctors strongly increased the proportion of mature protein in these cells, demonstrating that these molecules are functional in CF iPSC-derived cholangiocytes. To determine whether the correctors rescued functional expression of F508del-CFTR, we subjected treated and untreated cysts to the forskolin and IBMX-induced swelling assay (Fig. 7c,d). For these nature biotechnology advance online publication
studies, the small-molecule potentiator VX-770 was added along with forskolin and IBMX to the media. CF cysts showed little swelling in the absence of the correctors. However, with addition of both correctors, cysts generated from patient C1 increased by ~2.18 ± 0.52-fold, whereas those from patient GM00997 increased by 1.64 ± 0.08-fold 24 h after induction with forskolin, IBMX and VX-770 (Fig. 7c,d and Supplementary Movies 3 and 4). Although swelling over 24 h may reflect activities in addition to that of CFTR, sensitivity of swelling to CFTRinh-172 and other modulators of CFTR (cAMP agonists and VX-770) supports a major role for CFTR in this function. These findings show that it is possible to model CFTR defects in iPSCderived cholangiocyte-like cells and to use these populations to test patient-specific response to CFTR modulatory compounds. DISCUSSION Progress in uncovering the causes of biliary disorders and developing new therapies has been hampered by inaccessibility of the target tissue in patients and a lack of experimental model systems. Our study demonstrates an approach to model human biliary development and disease through a directed differentiation protocol that efficiently generates functional cholangiocyte cells from hPSCs that self-organize into duct-like structures in vitro and ectopically in vivo. This protocol was designed to recapitulate key aspects of cholangiocyte development in vivo, including the generation of hepatoblasts, specification of this population to a cholangiocyte fate, and induction and maturation of epithelialized duct-like structures. Through co-culture of OP9 cells with hPSC-derived hepatoblasts, we were able to mimic the Jagged1-Notch2 interaction that regulates mouse cholangiocyte development9–12 and demonstrate that generation of the human cholangio cyte lineage is also dependent on NOTCH signaling. Cholangiocyte specification with the co-culture approach appears to be very efficient as >90% of the structures that formed in 3D-cultures consisted of cholangiocyte-like cells and displayed bile duct characteristics.
© 2015 Nature America, Inc. All rights reserved.
Articles Whereas Notch functions to specify the cholangiocyte lineage from the hepatoblast31, the Egf, Hgf and Tgfβ pathways play a role in the formation and maturation of the biliary ductal system in the developing mouse8. EGF and HGF are thought to be required for stimulation of hepatoblast proliferation32, whereas TGFβ signaling is required to induce differentiation of biliary cells from periportal hepatoblasts24,33. We found that these pathways also influence cholangiocyte and ductal development in hPSC differentiation cultures. Whereas EGF and TGFβ signaling promoted proliferation of the CK19+ population, addition of HGF with these factors initiated formation of branched structures in two-dimensional (2D) monolayers. In 3D co-cultures, these factors promoted development of ductal and cystic structures with lumens comprising cells that displayed morphologic and functional characteristics of mature biliary epithelial cells. The presence of functional CFTR protein in hPSC-derived epithelial cells within 3D duct and/or cyst structures is strong evidence that they represent physiologically relevant biliary cells, as this transporter is known to be expressed on the apical membrane of mature bile ducts34,35. Patients with CF often develop liver dysfunction due to biliary cirrhosis36. Although there are no approved targeted therapies to treat CF patients, two correctors,VX-809 and VX-661, and a potentiator, VX-770, are currently in clinical trials. These molecules restore mutant plasma membrane CFTR expression and function to
Directed differentiation of cholangiocytes from human pluripotent stem cells
© 2015 Nature America, Inc. All rights reserved.
Mina Ogawa1,2,11, Shinichiro Ogawa1,11, Christine E Bear3, Saumel Ahmadi3, Stephanie Chin3, Bin Li4, Markus Grompe4, Gordon Keller1,5,6,12, Binita M Kamath7,8,12 & Anand Ghanekar2,9,10,12 Although bile duct disorders are well-recognized causes of liver disease, the molecular and cellular events leading to biliary dysfunction are poorly understood. To enable modeling and drug discovery for biliary disease, we describe a protocol that achieves efficient differentiation of biliary epithelial cells (cholangiocytes) from human pluripotent stem cells (hPSCs) through delivery of developmentally relevant cues, including NOTCH signaling. Using three-dimensional culture, the protocol yields cystic and/or ductal structures that express mature biliary markers, including apical sodium-dependent bile acid transporter, secretin receptor, cilia and cystic fibrosis transmembrane conductance regulator (CFTR). We demonstrate that hPSC-derived cholangiocytes possess epithelial functions, including rhodamine efflux and CFTR-mediated fluid secretion. Furthermore, we show that functionally impaired hPSC-derived cholangiocytes from cystic fibrosis patients are rescued by CFTR correctors. These findings demonstrate that mature cholangiocytes can be differentiated from hPSCs and used for studies of biliary development and disease. Protocols to generate tissue-specific cell types from human embryonic stem cells (hESCs) and induced pluripotent stem cells (iPSCs) have facilitated the development of novel in vitro models of human development and disease that have opened new avenues for drug discovery and cell therapy. Of the different lineages that can be studied with this approach, those of the liver are of special interest as hepatocytes and cholangiocytes, which make up the liver’s biliary system, are primary targets of the adverse effects of drugs and inherited and infectious diseases1. Biliary disorders such as cystic fibrosis (CF) and Alagille syndrome are common causes of liver disease that often necessitate liver transplantation2. More complex biliary diseases, such as primary sclerosing cholangitis and biliary atresia, lack appropriate models for understanding their pathophysiology and for screening pharmacological agents. Most studies to date have focused on generating hepatocytes from hPSCs3–6 and have yielded protocols that give rise to relatively mature cells suitable for predictive drug toxicology and modeling certain liver diseases in vitro5,6. Recently, cells with cholangiocyte characteristics were generated from hPSCs7, demonstrating that it is possible to derive this lineage in vitro as well. However, because these populations were established as monolayers, their functional capacity remains unclear, as cells cultured in this format often do not accurately recapitulate the normal physiological responses of complex tissues. The generation of functional three-dimensional (3D) biliary structures from hPSCs by following developmentally relevant signaling cues would overcome this limitation and provide an in vitro model
for studying biliary diseases and a foundation for developing tissue replacement strategies. Cholangiocytes derive from a bipotential progenitor known as the hepatoblast that also gives rise to hepatocytes8. Targeting studies in mice have shown that cholangiocyte specification from the hepatoblast is mediated by the interaction of Notch2, expressed by the progenitors, and Jagged1, present on the developing portal mesenchyme9,10. Notch signaling also plays a critical role in intrahepatic bile duct morphogenesis in mice11,12. The discovery that the pediatric biliary disorder Alagille syndrome, characterized by bile duct paucity, is caused by mutations in JAG1 or NOTCH2 provides strong evidence that NOTCH signaling is also involved in cholangiocyte development in humans13–15. However, the role of the NOTCH pathway in regulating human cholangiocyte development from hPSCs has not been investigated. Here we developed a serum-free protocol that directs the efficient differentiation and maturation of functional cholangiocytes from hPSCs by delivering developmentally relevant cues, including NOTCH signaling through co-culture with OP9 stromal cells, at the hepatoblast stage of development. We show that hPSC-derived cholangiocytes grown in 3D culture form epithelialized cystic and/or ductal structures that express markers found in mature bile ducts, including the CFTR. Cholangiocytes from normal hPSCs generate cystic and/or ductal structures and display epithelial functions, including CFTR-mediated fluid secretion, observed through the regulation of cyst swelling, after stimulation of the cAMP pathway with forskolin. Cysts generated
1McEwen
Centre for Regenerative Medicine, University Health Network, Toronto, Ontario, Canada. 2Toronto General Research Institute, University Health Network, Toronto, Ontario, Canada. 3Program in Molecular Structure & Function, The Hospital for Sick Children Research Institute, Toronto, Ontario, Canada. 4Department of Pediatrics, Oregon Health and Science University, Portland, Oregon, USA. 5Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada. 6Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada. 7Division of Gastroenterology, Hepatology and Nutrition, The Hospital for Sick Children, Toronto, Ontario, Canada. 8Department of Pediatrics, University of Toronto, Toronto, Ontario, Canada. 9Division of General Surgery, University Health Network, Toronto, Ontario, Canada. 10Department of Surgery, University of Toronto, Toronto, Ontario, Canada. 11These authors contributed equally to this work. 12These authors jointly directed this work. Correspondence should be addressed to A.G. ([email protected]) or S.O. ([email protected]). Received 2 February; accepted 19 June; published online 13 July 2015; doi:10.1038/nbt.3294
nature biotechnology advance online publication
Articles
RESULTS Characterization of hPCS-derived hepatoblasts To generate hepatoblasts from hPSCs, we used our previously published protocol6 (Fig. 1a) with the modification that the endoderm induction step was carried out in monolayers rather than in three-dimensional (3D) embryoid bodies. With this approach, populations with >90% of cells expressing the definitive endoderm markers CXCR4, cKIT and EPCAM (ref. 16) were generated by day 7 of differentiation (Fig. 1b). Hepatic specification of this endoderm population was achieved by the addition of bFGF and BMP4 (refs. 6,16,17). Formation of the mouse embryonic liver bud is dependent on transient expression of Tbx3 (refs. 18,19). As the liver bud expands, progenitor cells downregulate Tbx3 and maintain and/or upregulate the expression of genes normally expressed in the hepatic and/or cholangiocyte lineages, including albumin (Alb), alpha fetoprotein (Afp)
Day 3 Day 5 Day 7
104 103 0.123
0102 103 104 105
cKIT 105
0.188
99.6
4
10
103 102 0
0.047
0.212
0
0
8 6
AFP
4 2 0
e
AFP/TBX3
f % of max.
100 80
% of max.
1.0
2
20
1
0.5
0
0
0
ALB
CK19
Merge
AFP 61.6
100 60
11.7
40 20
96
40 20 0
0 102 103 104 105 ALB
CK19
80
100
19.4
40 20 0
0 102 103 104 105
NOTCH2
0 102 103 104 105
AFP
100
80 60
96.6
40 20 0
0 102 103 104 105
80 60
100
100 80 60
CK19
80 60 40
94.1
20 0
0 102 103 104 105
0 102 103 104 105
ALB
100 90.4
40 20 0
**
1.5 NOTCH1
DAPI
Day 13
***
Merge
80
0
NOTCH2
AFP
20
60
3
ALB
40
100
4
*
DAPI
60
0
* *
5
40
**
2.5 ALB 2.0 1.5 1.0 0.5 0
***
0
80 CK19 60
***
ES
EPCAM
d
5
10 HNF6
*** *** ***
10
10
C D s a D y7 ay D 13 ay D 19 ay 25 AL FL
0 102 103 104 105
20 SOX9 15
***
% of max.
0.617
20
% of max.
102 0
30 TBX3
HGF/EGF/ TGFβ +/– OP9
% of max.
97.3
Fold difference relative to fetal liver
CXCR4
1.95
HGF/Dex/OSM
bFGF/BMP4
0 102 103 104 105
% of max.
c
105
Day 19
% of max.
b
Day 13
ES C D s a D y7 ay D 13 ay D 19 ay 25 AL FL
Activin CHIR
NOTCH signaling promotes cholangiocyte development To investigate the effect of NOTCH signaling on cholangiocyte development, we co-cultured aggregates generated from day-25 hepatoblast populations with OP9 stromal cells, which are known to express different NOTCH ligands including JAGGED1 Cholangiocyte specification (refs. 21–23). As shown by immunostaining Day 25 and flow cytometry, the majority of cells within
Hepatoblast differentiation
% of max.
Day 0
Hepatic specification
cytokeratin 19 (Ck19; also known as Krt19), SRY-BOX 9 (Sox9), hepatic nuclear factor 6 (Hnf6; also known as Onecut1) and Notch2 (refs. 8,20). Analyses of bFGF- and BMP4-treated hESC-derived endoderm revealed transient upregulation of TBX3 at day 13, identifying this time as the stage of hepatoblast specification (Fig. 1c). The majority of cells at day 13 expressed TBX3, indicating that hepatoblast specification was efficient (Fig. 1d). The onset of SOX9 and HNF6 expression overlapped with that of TBX3 (Fig. 1c). However, unlike that of TBX3, expression of these genes continued to increase until day 25, the final day of analysis. Expression of ALB and AFP was upregulated at day 19 and also increased at day 25. CK19 showed a biphasic pattern, with peak levels detected at days 13 and 25. Immunofluorescent staining (Fig. 1e) and flow cytometry (Fig. 1f) indicated that the majority of cells at day 25 were ALB+, AFP+ and CK19+. These findings strongly suggest that the day-25 cell populations represent the expanded hepatoblast stage of development, equivalent to the liver bud stage in vivo. NOTCH2 but not NOTCH1 expression was also upregulated at day 25, further demonstrating that this population contains hepatoblasts capable of signaling through this pathway.
ES C D s a D y7 ay D 13 ay D 19 ay 25 AL FL
Definitive endoderm monolayer
ES C D s D ay 7 ay D 13 ay D 19 ay 25 AL FL
a
CXCR4
© 2015 Nature America, Inc. All rights reserved.
from CF patient iPSCs show impairment in forskolin-induced swelling that is rescued by addition of clinically relevant CFTR modulators. These findings demonstrate that delivery of NOTCH signaling to hPSC-derived hepatoblasts drives the generation of structurally and functionally mature cholangiocytes that can be used to model clinically relevant biliary diseases in vitro.
NOTCH2
80 60
95.7
40 20 0
Day 25
0 102 103 104 105
Figure 1 Characterization of the hepatoblast stage of development in hPSC differentiation cultures. (a) Schematic representation of the differentiation protocol. (GSK3 inhibitor: CHIR99021), bFGF (basic fibroblast growth factor), BMP4, HGF (hepatocyte growth factor), Dex (dexamethasone), OSM (oncostatin M), EGF (epidermal growth factor), TGFβ (transforming growth factor beta), OP9 (OP9 stromal cells). (b) Flow cytometric analyses showing the proportion of CXCR4+, cKIT+ and EPCAM+ cells in day-7 populations. (c) RT-qPCR analyses of the expression of the indicated genes in populations at different stages generated from H9 hESCs as indicated in Figure 1a. Values are determined relative to TATA-binding protein (TBP) and presented as fold change relative to the expression in fetal liver, which is set as 1. AL, adult liver; FL, fetal liver. ANOVA with Tukey’s HSD test compared with day-7 endoderm: *P < 0.05, **P < 0.01, ***P < 0.001. Error bars, s.d. ± the mean of three independent experiments. (d) Immunostaining analyses showing the proportion of AFP+ and TBX3+ cells in the differentiating population at day 13 of culture. TBX3, Alexa 488 (green); AFP Cy3 (red). Scale bar, 100 µm. (e) Immunostaining analysis showing the proportion of ALB+, AFP+ and CK19+ cells in the differentiating population at day 25 of culture. ALB, Alexa 488 (green); AFP and CK19, Cy3 (red); nuclei, DAPI (blue). Scale bars, 100 µm. (f) Intracellular flow cytometric analysis showing the proportion of AFP+, ALB+, CK19+ and NOTCH2+ cells in the populations at days 13 and 25 of culture. Blue, positive cells; red, isotype controls.
advance online publication nature biotechnology
Articles
DAPI
DAPI
b ALB
ALB
AFP
CK19
Merge
Merge
CTRL
EGF
c
e
EGF TGFβ1
HGF EGF
5
SOX9
CK19
Fold difference relative to OP9 (–)
a
3
HES5
HEY1
**
4
**
**
***
2 1 0
HGF
HES1
HGF EGF TGFβ1
OP9 OP9 OP9 OP9 OP9 OP9 OP9 OP9 OP9 (–) (+) (+) (–) (+) (+) (–) (+) (+) GSI GSI GSI (+) (+) (+)
Figure 2 NOTCH signaling promotes CFTR DAPI cholangiocyte differentiation from hPSC-derived hepatoblasts. (a) Confocal 2 ALB microscopic images of immunostained HGF EGF TGFβ1 day-27 aggregates showing co-expression TGFβ1 1 of ALB+, AFP+ and CK19+ cells. ALB, GSI Alexa 488 (green); AFP and CK19, 0 ASBT DAPI Cy3 (red); nuclei, DAPI (blue). Scale bars, 15 CK19 ALB DAPI CK19 50 µm. (b) Immunostaining analysis showing 10 the presence of ALB+ and CK19+ cells in hepatoblast populations co-cultured α-tubulin 5 CK19 for 9 d (total culture time 36 d) on OP9 stromal cells in the indicated conditions. DAPI CTRL: control with no additional cytokines. ALB, Alexa 488 (green); CK19 Cy3 0 (red); nuclei, DAPI (blue). Scale bars, 100 µm. (c) Immunostaining analysis + + + showing the presence of CK19 /SOX9 cells (upper panel), CFTR cells (middle panel) and ASBT+ cells (lower panel) in hepatoblast populations co-cultured for 9 d (total culture time 36 d) on OP9 stromal cells in HGF, EGF and TGFβ1. SOX9, Alexa 488 (green); CK19, CFTR and ASBT, Cy3 (red); nuclei, DAPI (blue). Scale bars, 100 µm. (d) Confocal microscopy images for immunostaining of acetylated α-tubulin localized in primary cilia on H9-derived cholangiocytes cultured on OP9 stroma in HGF, EGF and TGFβ1. Acetylated α-tubulin, Cy3 (red); nuclei, DAPI (blue). Inset image with triple staining confirms colocalization of CK19 and primary cilia. CK19, Alexa 488 (green). Scale bar, 10 µm. (e) RT-qPCR-based analysis of the expression of NOTCH target genes in hepatoblast populations after 9 d of culture (total 36 d) on OP9 stroma (OP9+) with or without GSI or Matrigel-covered wells (OP9 −) in the presence of HGF, EGF and TGFβ1. Values were determined relative to TBP and presented as fold-change relative to the levels in the cells cultured on Matrigel for 9 d (OP9 −), which is set as 1. Bars in all graphs represent the s.d. ± the mean of three independent experiments. Student’s t-test compared with the culture on OP9 stroma (OP9 +): **P < 0.01, ***P < 0.001 (n = 3). (f) RT-qPCR-based expression analysis of ALB and CK19 in hepatoblast populations cultured on OP9 stroma in the presence of HGF, EGF and TGFβ1 for 3 d (30 d total), 6 d (33 d total) and 9 d (36 d total). The addition of GSI is indicated. Hepatoblast (HB: day 27) represents the population before co-culture. Values were determined relative to TBP and presented as fold change relative to levels in the day 27 hepatoblast aggregates, which is set as 1. H Da B ( y d Da 30 ay 2 y G 7 Da 33 SI ) y G (– Da 36 SI ) y G (–) Da 30 SI y G (– Da 33 SI ) y G (+) 36 S G I(+) SI (+ ) AL FL
© 2015 Nature America, Inc. All rights reserved.
d
Fold change relative to hepatoblast
f
the aggregates before co-culture were ALB+AFP+CK19+NOTCH2+, indicating that they maintained hepatoblast characteristics (Fig. 2a and Supplementary Fig. 1a). When co-cultured on OP9 cells for 9 d, the aggregates formed distinct clusters of CK19+ cells that downregulated ALB expression, suggesting that they had undergone the initial stage of cholangiocyte specification (Fig. 2b, CTRL). As studies in mice have shown that Hgf, Egf and Tgfβ1 signaling plays a role in bile duct development8,20,24, we next added combinations of these factors to the cultures to determine whether these pathways would promote further development of CK19+ clusters. The addition of EGF alone, TGFβ1 alone or the combination of EGF, TGFβ1 and HGF led to a 3.0- to 3.5-fold increase in cell number compared with nontreated controls after 9 d of culture (Fig. 2b and Supplementary Fig. 1b). HGF alone had little effect. The majority (90%) of the cells in the populations cultured under these conditions were CK19+ and ALB−, indicating efficient specification and development of the cholangio cyte lineage (Supplementary Fig. 1c). The combination of either EGF and HGF or EGF, HGF and TGFβ1 induced substantial morphological changes and promoted the formation of CK19+ branched structures (Fig. 2b). CK19+ cells in structures treated with all three factors expressed other proteins indicative of the cholangiocyte lineage, including SOX9, the CFTR and the apical sodium-dependent bile acid transporter (ASBT) (Fig. 2c). They also displayed primary cilia as detected with acetylated α-tubulin (Fig. 2d). CFTR, ASBT and α-tubulin were not expressed in aggregates before OP9 cell co-culture (Supplementary Fig. 2). These findings demonstrate that co-culture of hepatoblasts with OP9 cells in the presence nature biotechnology advance online publication
of EGF, HGF and TGFβ1 promotes the development of a population that expresses cholangiocyte markers. With the efficiency of lineage specification in the co-culture model, we generated, on average, 4.16 ± 0.64 CK19+ALB− cholangiocyte-like cells per day-25 hepato blast and 6.42 ± 0.53 cholangiocyte-like cells per input hESC. Expression of NOTCH targets HES1, HES5 and HEY1 was upregulated after 9 d of culture on OP9 cells in the presence of EGF, HGF and TGFβ1 (Fig. 2e). This increase in expression was blocked by gamma secretase inhibitor (GSI), a NOTCH pathway antagonist, demonstrating that co-culture with OP9 cells effectively activated the NOTCH pathway. GSI also blocked downregulation of ALB, reduced the proportion of CK19+ cells and inhibited the development of branched structures in the cultures, indicating that these effects were mediated by NOTCH signaling (Fig. 2b and Supplementary Fig. 1c). RT-qPCR analyses (Fig. 2f) confirmed the immunostaining and flow cytometric findings and demonstrated a dramatic reduction in ALB and an upregulation of CK19 after co-culture with OP9 cells in the presence of EGF, HGF and TGFβ1. These expression patterns were reversed with GSI. Collectively, these findings indicate that activation of NOTCH signaling in the hepatoblast population induces the initial stages of cholangiocyte development and the combination of HGF, EGF and TGFβ1 signaling promotes the formation of branched structures reminiscent of early duct morphogenesis. Differentiation of two other hPSC lines (ESC HES2 and MSC-iPSC1) revealed similar temporal patterns of TBX3 and hepatoblast marker expression (Supplementary Fig. 3a). Aggregates of hepatoblasts from both generated branched structures consisting of CK19 +ALB− cells
Articles b Mix
Cyst
Tubular
Sphere
c 100
% of each type of structure
a
d
90 80 70 60 50 40 30 20 10 0
Mix Cyst Tubular Sphere
3D GSI(–)
3D GSI(+)
DAPI
CK19
ZO-1
Merge
DAPI
CK19
CFTR
Merge
© 2015 Nature America, Inc. All rights reserved.
GSI(–)
Figure 3 3D culture promotes cholangiocyte maturation. (a) Photomicrographs of duct-like and cyst-like structures that develop in the collagen/Matrigel cultures after 17 d of culture of chimeric aggregates consisting of H9-derived hepatoblasts and OP9 stromal cells (4:1 ratio) in the absence of GSI. Spheres DAPI SCTR CK19 Merge are dense colonies that develop in the presence of GSI. Scale bar, 100 µm. (b) Low-magnification photomicrograph showing cyst-like structures that develop from chimeric aggregates consisting of H9-derived hepatoblasts and OP9 stromal cells (4:1 ratio) after 17 d of culture in the collagen/Matrigel cultures. Scale bar, 100 µm. (c) Proportion of structures with tubular, cystic, DAPI ASBT CK19 Merge mixed and sphere morphology that developed per well from chimeric aggregates cultured in the absence or presence of GSI. Values represent averages from five DAPI ALB CK19 Merge independent experiments. (d) Confocal microscopy images of immunostained histological sections of 3D ductal and/or cyst structures after 17 d (total 44 d) GSI(+) of culture in the presence of HGF, EGF and TGFβ1 without GSI (GSI(−)). First and second rows: CK19, Alexa 488 (green); ZO-1 and CFTR, Cy3 (red). Third and fourth rows: SCTR and ASBT, Alexa 488 (green); CK19, Cy3 (red). Bottom ES OP9 OP9 row: confocal microscopic images of immunostained histological sections of 3D (kD) cells (+) (–) CK19 sphere structures after 17 d (total 44 d) of culture in the presence of HGF, EGF α-tubulin 250 and TGFβ1 with GSI (GSI(+)); ALB, Alexa 488 (green); CK19, Cy3 (red). In all DAPI B and C images, the nuclei are stained by DAPI (blue). Scale bars, 50 µm. (e) Confocal microscopy images of immunostaining of acetylated α-tubulin localized in α CFTR primary cilia present in a 3D ductal structure after 17 d (total 44 d) of co-culture 130 with OP9 stromal cells in the presence of HGF, EGF and TGFβ1. Acetylated α-tubulin, Alexa 488 (green); CK19, Cy3 (red); nuclei, DAPI (blue). Scale bar, 95 α CNX 10 µm. (f) Western blot analysis shows the presence of the mature glycosylated form of the CFTR protein (band C) in MSC-iPSC1-derived ductal and/or cyst structures generated after 17 d of culture of chimeric aggregates (OP9+) or aggregates of hepatoblasts without OP9 (OP9 −). Undifferentiated iPSCs were used as negative control and calnexin (CNX) was measured as a loading control.
e
following co-culture with OP9 stroma (Supplementary Fig. 3b), demonstrating the applicability of the approach to different hPSC lines. 3D culture promotes cholangiocyte maturation To determine whether growth in 3D structures would promote further maturation of the hPSC-derived cholangiocytes, we next generated chimeric aggregates consisting of day-25 hepatoblasts and OP9 (GFP +) stromal cells (Supplementary Fig. 4a) and cultured them in gels composed of 1.2 mg/ml collagen type-1, 40% Matrigel and containing HGF, EGF and TGFβ1. Within 2 weeks of culture, the aggregates gave rise to 3D structures that displayed either a tubular and ductal morphology, a hollow cyst morphology or a mixture of both (Fig. 3a–c). Expression of NOTCH target genes HES1, HES5 and HEY1 was significantly higher in structures derived from chimeric aggregates than in those aggregates without OP9 cells, indicating that NOTCH signaling was active under these conditions (Supplementary Fig. 4b). Approximately 17 (17.1 ± 6.06) structures developed per 105 day-25 input hepatoblasts, the majority (62.9 ± 30.7%) displaying the cyst morphology. The generation of these cyst and duct structures was dependent on NOTCH signaling, as they did not develop in the presence of GSI. Rather, under these conditions, we observed the formation of dense colonies that we refer to as spheres. Histological analyses revealed that tubular and cystic structures at day 44 had a lumen and contained epithelial-like cells that expressed
f
CK19 and E-CADHERIN but not ALB (Fig. 3d and Supplementary Fig. 4c). These cells also expressed the tight junction marker ZO-1 (Zonula occludens 1) CFTR, SCTR (secretin receptor) and ASBT and contained primary cilia (Fig. 3d,e). Most of these markers were restricted to the apical side of the structures, suggesting that the cells acquired apicobasal polarity, a feature of mature epithelial ducts. ASBT was also detected on the apical side of branching 3D structures (Supplementary Fig. 4d). The presence of only small numbers of Ki-67+ cells in CK19+ duct structures at this stage suggests limited proliferative capacity (Supplementary Fig. 4e). Similar patterns of CK19, CFTR and α-tubulin staining were observed in bile ducts of normal adult liver (Supplementary Fig. 5). Western blot analysis, RT-qPCR and flow cytometric analyses confirmed the presence of CFTR in cells generated from chimeric aggregates (Fig. 3f and Supplementary Fig. 6a,b). The levels of CFTR mRNA and protein were considerably higher in cells derived from chimeric aggregates than in those cultured without OP9 cells, indicating that CFTR expression was NOTCH-dependent. The spheres that developed in the presence of GSI expressed ALB and low levels of CK19, indicating that, in the absence of NOTCH signaling, cells with hepatoblast characteristics persist (Fig. 3d). RT-qPCR analyses showed similar changes as expression levels of hepatocyte markers (ALB, AFP and CYP3A7) increased whereas cholangiocyte-associated markers (CK19, SOX9 and CFTR) decreased (Supplementary Fig. 6c) after addition of GSI. advance online publication nature biotechnology
2.0 SOX9 1.5
1.5
*** ***
1.0
**
0.5
0.5 0 20
1.5
HNF1B
1.0
**
**
***
3
SCTR
2
10
** *
0 1.5 1.0
3
**
*
**
***
0 6
ASBT
*
**
4 2
0
15
*
5
1
0
1.5 TGR5
*
GGT
CFTR
20
2
0.5 0 20
0 25
10
0 4 SLC4A2/AE2
AQP1
1.0
15
1
5
HNF6
0.5
0 CK19
15
**
*
1.0
10 0.5
5
2.5
SSTR2
**
2.0 1.5
**
**
1.0 0.5
**
hE
C
H
hPSC-derived cholangiocytes form ductal structures in vivo To evaluate the developmental potential of hPSC-derived cholangiocytes in vivo, we transplanted 106 cells in a Matrigel plug into mammary fat pads of nonobese diabetic–severe combined immunodeficient–interleukin 2rγ−/− (NOD-SCID-IL2rγ−/−; NSG) mice. For these studies, we used dissociated cells from branched structures generated by co-culture of day-25 hepatoblasts derived nature biotechnology advance online publication
0 100
ALB
***
**
6,000 4,000
50
2,000 0
ep S ep ato i C at bla PS o h b s C C ola las t (H ho n t 9 la gio (iP ) n S Pr gio cyt C) c e i So ma yte (H rte ry ( 9) d hep iPS ch a C ol to ) an cy gi te oc yt e
To further validate the utility of the 3D 1 protocol, we next applied it to the human iPSC line MSC-iPSC1. A distribution of 0 CK19+/CFTR+/ASBT+ tubular and/or cystic structures similar to that observed in the H9 cultures developed from MSC-iPSC1-derived hepatoblast aggregates (Supplementary Fig. 7a). The efficiency of tubular and/or duct/cyst formation was also similar to that of the H9 hESCs, with 105 day-25 iPSC-derived hepatoblasts giving rise to 20.0 ± 1.64 structures. Primary cilia were also detected in these duct structures (Supplementary Fig. 7b). RT-qPCR analyses revealed that expression of genes encoding proteins found in functional cholangiocytes in vivo, including SOX9, HNF1B, HNF6, CK19, CFTR, SCTR, aquaporin-1 (AQP1), the Cl−/HCO3− anion exchanger 2 (SLC4A2/AE2), ASBT, gammaglutamyltransferase (GGT), TGR5 and the somatostatin receptor 2 (SSTR2) was significantly upregulated in hPSC-derived cystic structures compared to day-25 hepatoblasts (Fig. 4). Notably, the levels found in these structures were equivalent to or greater than those detected in primary cholangiocytes isolated by fluorescenceactivated cell sorting (FACS) from human liver using the ductal antibody DHIC5-4D9 (ref. 25; Supplementary Figs. 5 and 8). In contrast, HNF4A and ALB were expressed at higher levels in primary cultured hepatocytes and hPSC-derived hepatoblasts than in the cysts or primary cholangiocytes (Fig. 4 and Supplementary Table 3). Together, these findings show that when cultured in a mixture of Matrigel and collagen, the hepatoblast population can generate duct-like structures that express mature biliary markers.
***
***
0
hE H ep S ep ato i C P b a C to las SC h b C ola las t (H ho n t la gio (iP 9) n Pr gio cyt SC e ) i So ma cyte (H rte ry (iP 9) h d ep S ch a C ol to ) an cy gi te oc yt e
**
2
0 10,000 AFP 8,000
H
HNF4A
hE H ep S ep ato i C P b a C to las SC h b C ola las t (H ho n t la gio (iP 9) n Pr gio cyt SC e ) i So ma cyte (H rte ry (iP 9) h d ep S ch a C ol to ) an cy gi te oc yt e
0 3
H
Fold difference relative to sorted cholangiocyte
Figure 4 Gene expression profile of hPSCsderived cholangiocytes compared to sorted cholangiocytes. RT-qPCR analyses of the expression of the indicated genes in the indicated populations. hESC: undifferentiated H9 cells; iPSC: undifferentiated MSCiPSCs; hepatoblasts: day-25 populations; cholangiocytes: cystic structures generated from chimeric aggregates of either H9 or MSC-iPSCs-derived hepatoblasts; primary cultured hepatocytes; sorted cholangiocytes: cholangiocytes isolated from human liver samples by FACS, using the ductal-specific monoclonal antibody DHIC5-4D9 (ref. 25). Data are expressed as the fold change relative to sorted human cholangiocytes, which is set as 1. Student’s t-test: *P < 0.05, **P < 0.01, ***P < 0.001 (n = 3). Error bars in all graphs represent the s.d. ± the mean of three independent experiments. SCTR (secretin receptor), AQP1 (aquaporin 1), SLC4A2/AE2 (Anion exchange protein 2), ASBT, GGT (gamma-glutamyl transpeptidase), TGR5 (G protein- coupled bile acid receptor 1), SSTR2 (somatostatin receptor 2), HNF4A, AFP, (α-fetoprotein), ALB.
H
© 2015 Nature America, Inc. All rights reserved.
Articles
from HES2-RFP hESCs26 with OP9 stroma. Six to 8 weeks after transplantation, multiple duct-like structures were detected in the Matrigel plug (Supplementary Fig. 9a). Cells within the ducts were RFP+, demonstrating that they were of human origin (Supplementary Fig. 9b,c). Additionally they expressed CK19, CFTR and primary cilia, indicating that they displayed characteristics of cholangiocytes (Supplementary Fig. 9d,e). Teratomas were not observed in any of the animals receiving transplants. hPSC-derived cholangiocytes are functional To assess function of the hPSC-derived cholangiocytes, we evaluated their ability to efflux rhodamine 123, a dye used to measure activity of the multidrug resistance protein 1 (MDR1) transporter in normal bile duct cells. The cystic structures derived from H9 hESCs and normal iPSCs transported dye to the luminal space, indicative of transporter activity (Fig. 5a). In the presence of 20 µM verapamil, an inhibitor of MDR1, rhodamine did not accumulate in the lumen, confirming that dye movement reflected active transport through MDR1. To demonstrate CFTR functional activity, we performed a forskolininduced swelling assay on cystic structures (Fig. 5b). With this assay, activation of the cAMP pathway by addition of forskolin/IBMX increases CFTR function, resulting in fluid transport and cyst swelling27 that can be visualized after staining with the cellpermeable fluorescent dye calcein green. Addition of forskolin/IBMX to the cultures induced a fold increase of 1.23 ± 0.09 and 1.40 ± 0.17 in the size of H9- and iPSC-derived cysts, respectively, when measured 1 h later. Addition of the CFTR inhibitor CFTRinh-172 blocked the forskolin/IBMX-induced swelling, indicating that the increase
Articles b
MSC-iPSC1
c
(+)
MSC-iPSC1 500
***
300
*
***
***
***
400
***
300
200
200
100
100
0 FSK/IBMX (+) CFTR inhibitor (–)
MSC-iPSC1-derived cholangiocytes
Verapamil
H9 400
0 (+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
(+) (–)
(+) (+)
Figure 5 Function of hPSC-derived cholangiocytes in vitro. (a) Phase contrast images showing uptake of rhodamine 123 dye in 17-d-old cysts generated from cholangiocytes derived from either H9 hESCs or MSC-iPSC1 cells. To demonstrate functional transport, cysts were incubated with verapamil for 30 min before the addition of rhodamine 123. Images were taken 10 min after the addition of rhodamine (green). Scale bars, 100 µm. (b,c) Forskolininduced swelling assay demonstrating functional CFTR activity in hPSC-derived cysts. (b) Representative epifluorescence microscopy images of hESC and MSC-iPSC1-derived cysts before (−) and 24 h after (+) stimulation with forskolin/IBMX (F/I). Cells were labeled with calcein green and images were taken 24 h after F/I stimulation. Scale bars, 500 µm. (c) Quantification of the degree of cyst swelling 24 h after F/I stimulation in the absence or presence of the CFTR inhibitor CFTR Inh-172 (30 µM). Cyst swelling was quantified using velocity imaging software. Values shown are percent swelling relative to the cyst size before stimulation. Analyses from three independent experiments are shown. Student’s t-test: *P < 0.05, ***P < 0.001 (n = 3).
in cyst size was CFTR-dependent (Supplementary Fig. 10). When measured 24 h later, the swelling of H9- and iPSC-derived cysts increased 2.09 ± 0.21- and 2.65 ± 0.31-fold, respectively (Fig. 5c and Supplementary Movies 1 and 2). This swelling was also blocked by addition of CFTRinh-172. These findings demonstrate that cells in hPSC-derived cyst and/or duct-like structures display properties of functional cholangiocytes.
CF patients carrying the common F508del mutation (C1-iPSCs and GM00997-iPSCs). Both patient-derived iPSC lines generated definitive endoderm and hepatoblasts with kinetics similar to that of normal iPSCs (Fig. 6a,b and data not shown). Although hepatoblast development was not altered, cyst formation from the CF-iPSCs was clearly impaired as only branched structures were observed in gels after 17 d of culture (Fig. 6c). Cyst formation could be induced by addition of forskolin for the first week of the 2-week culture period (Fig. 6d). However, even under these conditions many cysts that developed from patient iPSCs were not completely hollow and contained branched ductal structures (Fig. 6e). As with the wild-type
10
3
98.8
2
10 0 0.2
0
2 0 10 103 104 105
cKIT 105 0
99.7
104 103 2
10 0 0
0.28
2 3 4 5 0 10 10 10 10
EPCAM
* ***
6 TBX3*** 4
*
2 0
25 AFP 20 15 10 5 0 5 SOX9 4 3 2 1 0
60
*** *
c
CK19
*
*
40
–FSK
20
*** ***
0 8
NOTCH2
6
***
4 2
ay
D
d
*** ***
D
ay
19
ay
D
19
MSC-iPSC1 Inh (+) Inh (–)
100
100
80
80
60
60
40
40
20
20
0
0
+FSK
0
19
C1-iPSC FSK (–) FSK (+)
7 d 14 d 7 d 14 d
4
ALB
Cyst formation (%)
10
2.5 2.0 1.5 1.0 0.5 0
C
CXCR4
105 1.04
Fold difference relative to fetal liver
b
1iP D SC D ay 7 ay 13 D ay 25 AL FL C 1iP D SC a D y7 ay 13 D ay 25 AL FL C 1iP D SC D ay ay 7 13 D ay 25 AL FL
a
e
MSC-iPSC1
7 d 14 d 7 d 14 d
Analyses of cholangiocytes generated from CF patient iPSCs To demonstrate the utility of this system to model disease in vitro, we analyzed cyst formation from iPSCs generated from two different
CXCR4
© 2015 Nature America, Inc. All rights reserved.
(–)
H9-derived cholangiocytes
DMSO
H9
% swelling relative to noninduced cysts
a
C1-iPSC
Figure 6 Generation of cholangiocytes from CF patient iPSCs. (a) Flow cytometric analyses showing the proportion of CXCR4+, cKIT+ and EPCAM+ cells in day-7 monolayer populations generated from a CF-iPSC line (C1 CFTR del). (b) RT-qPCR-based analyses of the expression of the indicated genes in populations at different stages generated from C1-iPSC. Cells were cultured as indicated in Figure 1a. Values shown are relative to TBP and presented as fold change relative to expression in fetal liver, which is set as one. AL: adult liver, FL: fetal liver. ANOVA with Tukey’s HSD test compared with day-7 endoderm: *P < 0.05, ***P < 0.001 (n = 3). Bars in all graphs represent the s.d. ± the mean of three independent experiments. (c) Photomicrographs of branched and cyst-like structures that develop in the collagen/Matrigel cultures after 17 d of culture of chimeric aggregates generated with cholangiocytes from CFTR F508del iPSCs (CF-iPSCs: C1-iPSC). Images are shown for structures that developed in the absence or presence of forskolin (FSK, 10 µM). Forskolin was added for the first week of the 17-d culture period (right). Scale bars, 100 µm. (d) Quantification of the number of cyst structures that developed from CF-iPSCs (C1-iPSC)-derived cholangiocytes cultured in the absence and presence of forskolin and from wild-type iPSCs (MSC-iPSC1) cultured in the absence and presence of the CFTR inhibitor CFTR Inh-172 (30 µM). Cyst structures were counted at 7 d and 14 d of culture. Forskolin was added to the CF-iPSC-derived aggregates and the CFTR inhibitor was added to the iPSC-derived aggregates for the first week of the 17-d-culture period (left graph). Bars represent the s.d. of the mean of three independent experiments. (e) Histological analyses showing the structure of iPSC- and CF-iPSC-derived cysts after 17 d of culture in the collagen/Matrigel gels (H&E staining; scale bars, 100 µm). Forskolin was included for the first week of the 17-d culture period.
advance online publication nature biotechnology
Articles Figure 7 Functional analysis of cholangiocytes from CF patient iPSCs. (a) Confocal images of histological sections showing the (–) Correction (+) Correction presence of the mutant CFTR protein in 16-d-old cyst structures (MW) C1-iPSC GM00997 – HBE generated from CF-iPSC-derived cholangiocytes cultured as chimeric (kD) iPSC aggregates in the presence or absence of the chemical correctors 250 α CFTR C VX-809 (3 µM) and Corr-4a (5 µM). The correctors were added at 2 weeks of culture. Forty-eight hours after the addition of the B (+) Correction 130 A correctors, the structures were fixed and analyzed for the presence α CNX of CFTR. Lower left panel shows the presence CFTR protein (Cy3, red) 95 (–) (+) (–) (+) in the cysts treated with the correctors. Membrane localized CFTR was VX-809 + Corr-4a not detected in the nontreated cysts (upper left panel). Right panel DAPI CK19 CFTR shows co-expression of CFTR and CK19 in a treated cyst. CK19, GM00997-iPSC Alexa 488 (green); CFTR, Cy3 (red); nuclei, DAPI (blue). Scale bars, (–) (+) 250 *** *** *** 50 µm. Arrows indicate a perinuclear distribution pattern of CFTR GM00997-iPSC 200 protein. (b) Western blot analysis showing the presence of the mature 150 glycosylated (band C) and core glycosylated (band B) form of the CFTR 100 protein in the nontreated and VX-809/Corr-4a-treated 16-d-old cysts. 50 HBE were used as a positive control. (c) Representative epifluorescence 0 microscopy images of calcein-green-labeled cyst structures generated C1-iPSC from cholangiocytes derived from CF-iPSC from two different patients * *** *** C1-iPSC (C1-iPSC and GM00997-iPSC). Images shown are before (−) and 24 h 400 after (+) stimulation with forskolin, IBMX and the CFTR potentiator 300 VX-770 (1 µM). Chimeric aggregates were generated in collagen/Matrigel 200 gels in the presence of forskolin for the first week of the 16-d culture period. 100 The correctors were added to the 2-week-old cultures for 48 h. At this stage, the cultures were stimulated with the combination of F/I and the FSK/IBMX 0 (+) (+) (+) (+) (+) (+) plus VX-770 potentiator VX-770. Cysts were measured 24 h later. Scale bars, 500 µm. VX-809 + Corr-4a (+) (–) (+) (–) (+) (–) (d) Quantification of the degree of swelling of the cysts derived from two different CF-iPSC lines. Cysts were cultured in the presence and absence of correctors, as indicated. Cyst swelling was quantified using velocity imaging software. The total size of the cyst was normalized to that before F/I stimulation from each of three individual experiments. Student’s t-test: *P < 0.05, *** P < 0.001 (n = 3).
b
c
d
© 2015 Nature America, Inc. All rights reserved.
% swelling relative to noninduced cysts
a
cells, CF iPSC-derived cysts contained primary cilia (Supplementary Fig. 11). A higher frequency of hollow cysts was detected following longer periods of culture, suggesting that maturation of the mutant cells was delayed. Addition of CFTR inhibitors to cultures of normal iPSC-derived cholangiocytes also delayed cyst formation, indicating that generation of these structures was dependent, to some degree, on functional CFTR at the cell surface (Fig. 6d). We assessed functional restoration of F508del CFTR in cholangio cytes from CF-patient iPSCs by adding the chemical correctors VX-809 and Corr-4a to cultures 2 d before the CFTR functional assay27–30. Both molecules correct folding defects of mutant CFTR protein. The addition of these correctors did not improve cyst formation but did result in accumulation of detectable levels of CFTR on the apical side of the lumen. Unlike the homogeneous distribution of CFTR in wild-type cysts, the mutant protein localized to the plasma membrane in some areas and exhibited perinuclear distribution in other regions (Fig. 7a). This pattern may reflect incomplete rescue of the trafficking defect of the mutant protein. Western blot analyses also revealed the effect of correctors. Wild-type CFTR normally migrates as two bands, one ~130–150 kD (band B) and the other ~170–190 kD (band C). This typical CFTR protein migration pattern is shown for lysates from human bronchial epithelial (HBE) cells (Fig. 7b). Band C represents the mature, Golgi-modified CFTR and band B, the immature, ER-localized CFTR protein. The F508del-CFTR protein migrated as a 130-150 kD protein in the iPSC-derived cholangiocytes in the absence of correctors consistent with retention in the endoplasmic reticulum (ER) and failure to traffic to the cell surface. Treatment with correctors strongly increased the proportion of mature protein in these cells, demonstrating that these molecules are functional in CF iPSC-derived cholangiocytes. To determine whether the correctors rescued functional expression of F508del-CFTR, we subjected treated and untreated cysts to the forskolin and IBMX-induced swelling assay (Fig. 7c,d). For these nature biotechnology advance online publication
studies, the small-molecule potentiator VX-770 was added along with forskolin and IBMX to the media. CF cysts showed little swelling in the absence of the correctors. However, with addition of both correctors, cysts generated from patient C1 increased by ~2.18 ± 0.52-fold, whereas those from patient GM00997 increased by 1.64 ± 0.08-fold 24 h after induction with forskolin, IBMX and VX-770 (Fig. 7c,d and Supplementary Movies 3 and 4). Although swelling over 24 h may reflect activities in addition to that of CFTR, sensitivity of swelling to CFTRinh-172 and other modulators of CFTR (cAMP agonists and VX-770) supports a major role for CFTR in this function. These findings show that it is possible to model CFTR defects in iPSCderived cholangiocyte-like cells and to use these populations to test patient-specific response to CFTR modulatory compounds. DISCUSSION Progress in uncovering the causes of biliary disorders and developing new therapies has been hampered by inaccessibility of the target tissue in patients and a lack of experimental model systems. Our study demonstrates an approach to model human biliary development and disease through a directed differentiation protocol that efficiently generates functional cholangiocyte cells from hPSCs that self-organize into duct-like structures in vitro and ectopically in vivo. This protocol was designed to recapitulate key aspects of cholangiocyte development in vivo, including the generation of hepatoblasts, specification of this population to a cholangiocyte fate, and induction and maturation of epithelialized duct-like structures. Through co-culture of OP9 cells with hPSC-derived hepatoblasts, we were able to mimic the Jagged1-Notch2 interaction that regulates mouse cholangiocyte development9–12 and demonstrate that generation of the human cholangio cyte lineage is also dependent on NOTCH signaling. Cholangiocyte specification with the co-culture approach appears to be very efficient as >90% of the structures that formed in 3D-cultures consisted of cholangiocyte-like cells and displayed bile duct characteristics.
© 2015 Nature America, Inc. All rights reserved.
Articles Whereas Notch functions to specify the cholangiocyte lineage from the hepatoblast31, the Egf, Hgf and Tgfβ pathways play a role in the formation and maturation of the biliary ductal system in the developing mouse8. EGF and HGF are thought to be required for stimulation of hepatoblast proliferation32, whereas TGFβ signaling is required to induce differentiation of biliary cells from periportal hepatoblasts24,33. We found that these pathways also influence cholangiocyte and ductal development in hPSC differentiation cultures. Whereas EGF and TGFβ signaling promoted proliferation of the CK19+ population, addition of HGF with these factors initiated formation of branched structures in two-dimensional (2D) monolayers. In 3D co-cultures, these factors promoted development of ductal and cystic structures with lumens comprising cells that displayed morphologic and functional characteristics of mature biliary epithelial cells. The presence of functional CFTR protein in hPSC-derived epithelial cells within 3D duct and/or cyst structures is strong evidence that they represent physiologically relevant biliary cells, as this transporter is known to be expressed on the apical membrane of mature bile ducts34,35. Patients with CF often develop liver dysfunction due to biliary cirrhosis36. Although there are no approved targeted therapies to treat CF patients, two correctors,VX-809 and VX-661, and a potentiator, VX-770, are currently in clinical trials. These molecules restore mutant plasma membrane CFTR expression and function to
