Ductular morphogenesis and functional polarization of normal human biliary epithelial cells in three-dimensional culture

Journal of Hepatology 35 (2001) 2±9

www.elsevier.com/locate/jhep

Ductular morphogenesis and functional polarization of normal human biliary epithelial cells in three-dimensional culture Yuichi Ishida 1, Sharon Smith 2, Lorraine Wallace 2, Takaharu Sadamoto 1, Masashi Okamoto 1, Marcus Auth 3, Mario Strazzabosco 4, Luca Fabris 4, Juan Medina 5, JesuÂs Prieto 5, Alastair Strain 6, James Neuberger 1, Ruth Joplin 2,* 1 Liver Unit, University Hospital, Birmingham, UK Department of Medicine, University of Birmingham, Birmingham, UK 3 Zentrum fur Kinderheilkunde, Universitatsklinikum, Essen, Germany 4 Clinica Medica Interna, Universita di Padova, Padova, Italy 5 Department of Medicine and Liver Unit, ClõÂnica Universitaria, Navarra University School of Medicine, Pamplona, Spain 6 School of Biochemistry, University of Birmingham, Birmingham, UK 2

Background/Aims: The understanding of the physiology and function of human biliary epithelial cells (hBEC) has been improved by studies in monolayer culture systems. The aim was to develop a polarized model to elucidate the mechanisms of ductular morphogenesis and functional differentiation of hBEC. Methods: The morphological, phenotypic and functional properties of hBEC cultured as three-dimensional aggregates in collagen gel were assessed in medium supplemented with (or without) human hepatocyte growth factor (hHGF) and foetal bovine serum. Results: In the absence of added mitogens and serum, cells maintained as morphologically polarized aggregates, organized around a central lumen, were positive for phenotypic markers of biliary epithelium and negative for markers of other cell types. Functional markers, gamma-glutamyl-transferase, anion exchanger-2, responses to gamma interferon and forskolin induced secretion, were preserved. hHGF increased both the size and number of aggregates and induced hBEC to invade the gel and lumena forming anastomosing networks of cells. Conclusions: Collagen gel culture in the absence of added growth factors and serum provides a model for analysis of the polarized functions of hBEC. The formation of poorly organized cords of cells in response to hHGF suggests that collagen gel culture may provide a model for the investigation of atypical ductular morphogenesis of the human biliary tract. q 2001 European Association for the Study of the Liver. Published by Elsevier Science B.V. All rights reserved. Keywords: Hepatocyte growth factor; Collagen gel; Ductular morphogenesis; Biliary epithelial cells

1. Introduction The intrahepatic biliary tract is a target for damage in conditions including primary cholangiopathies and allogeneic reactions following transplantation. Human intrahepatic biliary epithelium is a polarized tissue that lines the Received 19 July 2000; received in revised form 15 January 2001; accepted 27 February 2001 * Corresponding author. Liver Research Laboratories, University Hospital, Edgbaston, Birmingham B15 2TH, UK. Tel.: 144-21-414-3917; fax: 144-21-627-2497. E-mail address: [email protected] (R. Joplin).

biliary tract. Despite representing ,5% of the total cells in normal liver [1], human intrahepatic biliary epithelial cells (hBEC) have been isolated with high purity for in vitro studies. Cultured hBEC have increased our understanding of their functional properties [2±9]. However, the interpretation of studies in vitro depends upon the in vitro model re¯ecting properties of cells in vivo. The maintenance of polarity, a well-differentiated phenotype and function are important considerations when culturing epithelial cells. Several techniques have been described for maintenance of hBEC in short- but not long-term tissue culture [10±12]. Previously, we showed that human hepato-

0168-8278/01/$20.00 q 2001 European Association for the Study of the Liver. Published by Elsevier Science B.V. All rights reserved. PII: S01 68-8278(01)0007 8-2

Y. Ishida et al. / Journal of Hepatology 35 (2001) 2±9

cyte growth factor (hHGF) is a mitogen for primary hBEC in monolayers [13]. However, monolayer culture is associated with morphological and phenotypic alterations of epithelia [14]. Furthermore, while monolayer culture permits access to the apical membrane of epithelia, the basal membrane is inaccessible. Culture of a polarized rodent cholangiocyte line on permeable supports permitted access to the basal membrane and assessment of basal and apical functions [15]. However, this technique cannot be used with primary hBEC because of the aggregate nature of freshly puri®ed cells [10±12]. Complex substrata improve the morphology and function of epithelia in vitro [16,17]. In gels of collagen, rodent biliary epithelial cells (rBEC) are polarized and form tubular structures enclosing a central lumen [18]. Similar improvements in the phenotypic and functional integrity may also be achieved in hBEC; hBEC co-cultured with hepatocytes in collagen, proliferated and formed hollow aggregates [19]. Here, we report that hBEC cultured in collagen maintain a differentiated phenotype, are morphologically polarized and functionally active. 2. Materials and methods 2.1. Preparation of collagen Type I collagen was obtained from rat-tail tendons (by modi®cation of established procedures, [19]). Collagen solution (1 ml/well) was pipetted into 6-well tissue culture plates and polymerized at 378C (20±30 min) [18± 20].

2.2. Isolation and culture of cells hBEC were puri®ed from normal adult human liver (n ˆ 12) using established methods [11,12]. Brie¯y, approximately 30 g of liver was digested with collagenase (type 1A), partially puri®ed by Percoll density centrifugation, and ®nally, highly puri®ed by immunomagnetic separation, using a monoclonal antibody (HEA125, Progen Biotechnik, Heidelberg, FRG) which recognizes a cell surface glycoprotein expressed solely in normal adult human liver by biliary epithelial cells [21]. This procedure provides hBEC of .95% viability and purity as determined by Trypan-Blue exclusion and immunocytochemistry using a panel of hBEC speci®c and nonspeci®c antibodies [11±13]. Puri®ed hBEC were adhered onto collagen gels (3 £ 10 3 cells/well) and 25 cm 2 tissue culture ¯asks (monolayer culture, 10 4 cells/ml), in medium enriched with foetal bovine serum (FBS) at 378C in a humidi®ed, 5% CO2 atmosphere [13]. In parallel experiments, hBEC were co-cultured in collagen gel with autologous hepatocytes, which, in previous studies, were shown to enhance the proliferation of hBEC [19]. Hepatocytes (.90% viability), prepared by collagenase perfusion as described previously [22], were plated onto collagen gels with hBEC to a ®nal density of 1.4±0.3% hBEC. In all experiments, the plating medium was discarded after 24±48 h. A second layer of collagen was layered over the hBEC and, following polymerization, fresh medium was added (serum-free Williams E, supplemented with 5% FBS and/or 10 ng/ml hHGF, Gibco; [19]). In monolayers, hBEC were cultured in our standard hBEC monolayer culture medium, which contains 5% FBS and 10 ng/ml hHGF [13].

2.3. Characterization of cultured cells Cells were examined daily by phase contrast microscopy. The number and size of cell aggregates were assessed in situ on days 4, 7, 10 and 14 of culture, using an eyepiece graticule. Cell polarity and invasion were

3

assessed using phase contrast microscopy, transmission electron microscopy (TEM) and immunocytochemistry. For TEM, cells were ®xed in situ and processed as described previously [19]. The phenotype was determined immunocytochemically using a panel of antibodies against various intermediate ®laments and cell surface markers (detailed in Table 1). Collagen gels were removed from the wells, snap frozen in liquid nitrogen, and 9 mm sections, cut at 2308C, were ®xed in acetone before immunostaining as described previously [23]. Antibody binding was visualized using an alkaline phosphatase/fast-red system, and sections were counter-stained with haematoxylin before analysis by laser scanning confocal microscopy. Assays for hBEC function included: (1), gamma-glutamyl-transferase (gGT) activity, detected histochemically [24] or by a soluble assay system (Boehringer Mannheim GmbH Diagnostica); (2), induction of intercellular adhesion molecule 1 (ICAM-1) and human leucoycte antigen (HLA) class II, determined using a soluble ELISA assay [25], by immunoblotting [9] or by immunocytochemistry after culture with or without gamma interferon (gIFN, 0.1±400 U/ml) for 3 days. For ELISA, 2 £ 10 4 hBEC/well were plated in 96-well plates. For PAGE and Western blotting, 10 mg hBEC protein/lane were loaded onto 10% polyacrylamide gels, and electrophoresis and Western blotting were carried our as described previously [9]. ELISA plates and Western blots were stained with anti-HLA class II (Dako, 1:100). For ELISA, antibody binding was detected using 1,2phenylenediamine dihydrochloride (OPD), and for Western blots, a peroxidase/diaminobenzidine technique was used. ELISA plates were read at 492 nm using a MRX-microplate reader (Dynatech). Blots were analyzed by densitometry scanning; (3), Secretory activity of hBEC in the presence of 3 mM forskolin (Sigma) [2] was calculated by measuring the percentage change in size of hBEC aggregates during video cinematography (0±180 min), i.e. the diameter of aggregate prior to forskolin stimulation/diameter of same aggregate following forskolin £ 100 (n ˆ 10). Table 1 Comparison between hBEC cultured as monolayers and in collagen a,b Characterization c

Collagen gel Monolayers

Morphological polarity Phenotypic markers HBEC CK-19 (%) Epithelial glycoprotein 34 (HEA125; %) EMA (%) Other Factor VIII related antigen (%) Vimentin (VMN; %) CD31 (%) Desmin (%) Fibroblast antigen (DIA 100; %) Functional markers gGT AE2 ICAM induction MHC Cl II induction Urea production Albumin production Lidocaine metabolism

Yes

No

. 95 . 95

. 95 50 (approximately)

. 95

. 95

,2 ,2 ,2 No No

No 50 (approximately) No ,2 ,2

Yes Yes Yes Yes No No No

Yes Yes Yes Yes No No No

a

The number of positive cells following 2±4 weeks of culture are expressed as the percentage of total cells. b Cells in collagen gel were positive for hBEC markers at all time-points tested. HEA125 positivity was lost from hBEC cultured as monolayers after 2±4 weeks in culture. c All antibodies were obtained from Dako, except for HEA125 (Progen Biotechnik, GmbH), AE2 [45] and DIA 100 (Dionovo, GmbH), and were diluted at 1:100, except for EMA (1:20), AE2 (1:10) and ICAM-1 (1:250).

4

Y. Ishida et al. / Journal of Hepatology 35 (2001) 2±9

Table 2 Overview of results: effect of FBS and HGF on hBEC cultured as monolayers and in collagen Culture regime Collagen 5% FBS 10 ng/ml HGF 5% FBS/10 ng/ml HGF Control d Monolayers 5% FBS 10 ng/ml HGF 5% FBS/10 ng/ml HGF Control d

Growth a

Migration/ invasion b

Morphology c

No Yes Yes No

No Yes Yes No

Yes No No Yes

No No Yes No

No No Yes No

No No No No

a

Growth determined by increase in cell number and/or aggregate size/ number. b Migration/invasion determined by cell scattering in monolayers, or invasion of the lumen and extracellular matrix in collagen gels. c Morphology de®ned as maintenance of polarity, with in vitro duct formation (de®ned as a single layer of adjacent polarized BEC surrounding a discernible lumen). d Controls consisted of cells cultured in the absence of FBS and HGF.

Hepatocyte function was determined by: (1), measuring the daily albumin production using an immunoturbidimetric assay. Rabbit anti-human albumin (Dako) was detected and determined by COBAS MIRAS (Hoffmann La Roche); (2), urea production using a quantitative colorimetric

method (Sigma Diagnostics Urea Nitrogen) and a CE 292 digital ultraviolet spectrophotometer (Cecil instruments Ltd.) at 535 nm; (3), lidocaine metabolism assessed by assaying for its primary metabolite, monoethylglycinexylidide (MEGX). After incubation for 1 h with 50 mM/l lidocaine, samples were analyzed by quantitative measurement of MEGX using a ¯uorescence polarization immunoassay with a TDx analyzer (Abbott, IL).

3. Results A summary of results is given in Tables 1 and 2. 3.1. Morphological polarity, phenotype and function of hBEC are preserved in collagen gel culture Cells plated onto collagen retained the `ovoid' aggregate morphology of freshly isolated cells (Fig. 1a±b), while on plastic, the cells spread, forming ¯attened `islands'. Despite attachment of hBEC to plastic, the cells deteriorated and detached during the subsequent 7±14 days. By contrast, hBEC in collagen became organized into hollow, ovoid structures composed of adjacent hBEC surrounding a central lumen (Fig. 1c±f). The hBEC were morphologically polarized (having basal nuclei, apical microvilli and junctional complexes between adjacent cells; Fig. 1c±d), stained positively for hBEC speci®c markers and were negative for all other markers tested (Table 1; Fig. 1e±f). Cells could be maintained in this con®guration for up to 6 weeks.

Fig. 1. (a,b) Phase contrast images; (c,d), transmission electron micrographs; and (e,f), laser scanning confocal micrographs of: (a), hBEC freshly isolated; and cultured in collagen gel for: (b), 2; (e,f), 14; and (c,d), 28 days. (e,f) Cells stained by immunocytochemistry for HEA125. Note the morphological similarity between cells in: (b), collagen gel; and (a), freshly isolated aggregates, and the morphological polarization, with nuclei (N) situated toward the basal pole and apical microvilli (MV) and junctional complexes (arrow, d). (a) Arrows represent dynabeads. L, lumen.

Y. Ishida et al. / Journal of Hepatology 35 (2001) 2±9

5

3.2. HBEC in collagen gel form anastomosing networks in response to hHGF

Fig. 2. Number of hBEC aggregates observed/well in collagen gels in response to 10 ng/ml hHGF (solid line). Values are means ^ SEM. The broken line represents control cells cultured without hHGF. The decline in number of aggregates following 7 days of culture with hHGF was due to expansion and coalescence of adjacent aggregates and not because of a decrease in cell number.

An increase in the lumenal diameter in response to forskolin stimulation was observed in 70% of aggregates (range, 9±30% increase; median, 25%). ICAM-1 (not shown) and HLA class II were induced on hBEC in collagen by incubation with 100 U/ml gIFN for 3 days (Fig. 4d). Albumin, urea and MEGX were not detected (not shown).

Cell aggregates cultured in collagen gel in serum-free medium supplemented with 10 ng/ml hHGF expanded both in size and number during the ®rst 7 days of culture (Fig. 2). The aggregates developed irregular marginal projections that invaded the collagen and lumen, forming anastomosing networks (Fig. 3a±b). After 1±2 weeks, adjacent aggregates coalesced, and after 2±3 weeks, the cells became con¯uent. Ultrastructurally, these hBEC were ¯attened relative to freshly isolated hBEC and hBEC cultured in collagen without hHGF. Microvilli were less numerous and shorter, but some polarity was still apparent and junctional complexes could be observed between adjacent cells (Fig. 3c). Phenotypic positivity for markers of hBEC was retained. Attempts to release hBEC from the gels for subculture were unsuccessful. The use of high activity enzyme solutions (collagenase type XI) [19], while releasing the cells, proved damaging, and successful sub-culture of hBEC cultured in collagen gel could not be achieved. By contrast, despite little increase in cell number, hBEC aggregates cultured without hHGF were maintained for up to 6 weeks without loss of polarity or functional and phenotypic markers. Addition of FBS to the medium had no detectable effect on hBEC cultured in collagen gel.

Fig. 3. (a) Phase contrast; (b), immunocytochemical; and (c), TEM images of hBEC cultured in collagen gel with 10 ng/ml hHGF. Note: (a), the peripheral projections (arrow); and (b), anastomosing network of cells positive for HEA125 (N, nuclei stained with haematoxylin). The polarity of hBEC cultured in collagen gel in the presence of hHGF was reduced compared with those cultured without hHGF (compare with Fig. 1c,d), but microvilli and junctional complexes could still be observed (arrow, c).

6

Y. Ishida et al. / Journal of Hepatology 35 (2001) 2±9

3.3. HBEC can be maintained in medium-term monolayer culture in the presence of FBS and hHGF

(BEC) relative to hepatocyte only controls), but not at any other time point.

hBEC cultured in monolayers in the presence of 5% FBS and 10 ng/ml hHGF proliferated and formed con¯uent `cobblestone' monolayers of cells after 7±14 days. These cells could be passaged and maintained serially through 5±9 sub-cultures. The median survival was 14 weeks (range, 11±21 weeks) and a 10 5-fold increase in cell number was achieved following seven sub-cultures. Both FBS and hHGF were essential for this response. Ultrastructurally, hBEC in monolayer culture lost polarity within a few hours of plating, becoming ¯attened, with a central nucleus and sparse short microvilli. Cytokeratin-19 (CK-19) and epithelial membrane antigen (EMA) were maintained in .95% of cells after at least 15 weeks of culture. By contrast, HEA125 was expressed only during the ®rst 3±4 weeks of culture, and thereafter, was lost. Cells were uniformly negative for ®broblast antigen DIA 100, desmin, CD31 and Factor VIII related antigen at all time points. Vimentin was positive in ,2% of freshly isolated cells, but following 3±4 weeks of monolayer culture, the number of cells expressing vimentin increased, and after approximately 5 weeks, .95% of cells were positive for vimentin. A dose-dependent increase in ICAM 1 (not shown) and HLA class II, was observed when hBEC in monolayers were stimulated with gIFN (Fig. 4a±c). The magnitude of the response did not vary between cells that had been in culture for a range of time points between 3 and 11 weeks. AE2 and gGT were maintained in .95% of hBEC for at least 12 weeks of culture (not shown). The activity of gGT/10 6 cells (mean activity, 32 IU/l per 10 6 cells) did not vary signi®cantly between 3 (42 IU/l per 10 6 cells) and 15 weeks (30 IU/l per 10 6 cells) of culture. 3.4. Co-culture with autologous hepatocytes induces proliferation of hBEC but does not improve hepatocyte function In co-culture with hepatocytes, the number of hBEC aggregates as determined by morphological characteristics (Fig. 5a) increased during the ®rst 7 days of culture, and this was proportional to the number of hepatocytes plated (Fig. 5b). Thereafter, small aggregates coalesced to form larger ones, before deteriorating. hBEC aggregates were not observed when hepatocytes were cultured alone. Co-culture of hepatocytes with hBEC had no effect on albumin secretion by hepatocytes; albumin production by hepatocytes peaked after 7 days of culture and then declined. No difference in lidocaine metabolism was observed between cocultures and hepatocytes cultured alone. Urea production increased for the ®rst 4 days of culture and then declined. Co-culture of hepatocytes with hBEC resulted in increased urea production relative to hepatocytes cultured alone on days 4 and 7 (mean increase, 2.4:1, ratio of urea produced by hepatocytes co-cultured with biliary epithelial cells

Fig. 4. Dose-dependent induction of HLA class II on hBEC cultured as monolayers with gIFN for 3 days demonstrated by: (a), ELISA (mean absorbance ^ SEM); and (b), Western blotting (lanes 2±6, respectively: 0.1, 1, 10, 100, 200 U/ml gIFN; lane 7: no gIFN control; lanes 1 and 8: molecular weight markers). (c) Densitometry scanning of Western blots (mean absorbance ^ SEM). (d) Induction of HLA class II on hBEC cultured in collagen gel by incubation for 3 days with 100 U/ ml gIFN. L, lumen. Arrow represents dynabead.

Y. Ishida et al. / Journal of Hepatology 35 (2001) 2±9

Fig. 5. (a) Phase contrast micrograph showing the morphological appearance of hBEC co-cultured with autologous hepatocytes. Aggregates of hBEC (A) can be identi®ed between hepatocytes (H). (b) Quantitation of hBEC aggregates. The chart shows the mean number of hBEC aggregates observed/well when 3 £ 10 3 hBEC were co-cultured with a range of densities of autologous hepatocytes.

4. Discussion Recent studies increasingly demonstrate immunological and functional differences between rBEC and hBEC [2,7,17,26,27]. While studies in rodents suggest improved morphological, phenotypic and functional stability results when rBEC are cultured in collagen gel rather than as conventional monolayers [18], the optimum conditions for culture of hBEC are uncertain. Here, we studied normal hBEC cultured either as monolayers or in collagen. hBEC were maintained in gels for up to 6 weeks without loss of morphological polarity or phenotypic/functional attributes. In the absence of added mitogens, little growth was observed, but rapid proliferation resulted on addition of hHGF. Aggregates increased both in size (attributable to an increase in cell number within individual aggregates) and number (`new' aggregates observed after several days of culture, and likely to arise from the proliferation of cells in existing aggregates not recognized initially because of their small size). A decrease in the number of aggregates after 7 days of culture was attributable to the coalescence of adjacent aggregates as they expanded and did not represent a decline in cell number. Originally identi®ed as an epithelial cell mitogen [28], HGF is now recognized as an important morphogen capable of inducing arborization and tubule formation in a variety of cell types; it is also a homologue for scatter factor, a potent epithelial cell motogen [29±31]. Our data suggest that hBEC react to hHGF with all three biological responses. In collagen gels, invasive anastomosing cords of cells were observed in response to hHGF. These cords super®cially resembled atypical hyperplastic ductules observed in cirrho-

7

tic liver in vivo. A high level of ®brotic tissue and the release of factors by cells of mesenchymal origin in vivo are known to have a role in the migratory behaviour of epithelia. Our observations suggest that the incubation of hBEC with hHGF may recapitulate this response in vitro. Johnson et al. [32] demonstrated that 20 ng/ml HGF induced tubule formation in a normal rat non-parenchymal epithelial cellderived, cell line cultured in collagen gel. Blair et al. [33] demonstrated that trout BEC formed aggregates that showed shaft-like marginal projections, which by TEM appeared to represent biliary ductules forming in vitro. However, it is not clear whether these were well-formed ductular structures or disorganized, atypical ductular formations. Further studies are required to fully de®ne the relationship between growth and differentiation of hBEC and to characterize the cells and structures occurring in collagen gel in response to growth signals of epithelial and mesenchymal origin. Recently validated methods to segregate hBEC of atypical ductules from those of well-formed ducts and ductules, based on their different spectrum of phenotypic markers [34], may provide the opportunity to study ductular hyperplasia in vitro, including functions dependent on intercellular communication, lumen formation, polarization and arborization. In monolayers, hHGF induced a 10 4-fold increase in cell number after 3±4 weeks of culture. Phenotypically, hBEC at this stage were .95% positive for EMA and CK-19 and were uniformly negative for all control markers, with the exception of vimentin. Occasional vimentin positive cells seen in initial cell isolates and early cultures of hBEC (7 days) consistently failed to incorporate 3H-thymidine [13]. Similarly, the proportion of rBEC expressing vimentin increased only when the cells were cultured as monolayers on plastic [18]. When cultured in collagen gel, the number of vimentin positive rBEC remained stable at around 2%. It is known that vimentin expression is not con®ned exclusively to cells of mesenchymal origin. Vimentin expression by newly formed rBEC in vivo has been reported [35], demonstrating that epithelia have the capacity to express vimentin. Furthermore, vimentin expression is associated with invasive, migratory behaviour and metastasis. Collectively, the data suggest that in the present study, the large increase in vimentin positive cells after extended monolayer culture is unlikely to be due to overgrowth by contaminant cells, but is probably attributable to the induction of vimentin gene expression in hBEC by the culture conditions. The magnitude of gGT activity and the optimal concentration of gIFN for maximal induction of HLA class II and ICAM-1 (100±200 U/ml) following 3 months of culture were in close agreement with previous studies restricted to early primary cultures of BEC [36,37]. Furthermore, previous studies of hBEC in monolayers demonstrated the preservation of apical membrane functions, such as AE2 [2] and cystic ®brosis transductance regulator (CFTR) [38]. Thus, we have found that hBEC expanded in monolayer culture in the presence of hHGF exhibit considerable phenotypic and

8

Y. Ishida et al. / Journal of Hepatology 35 (2001) 2±9

functional stability. Despite the changes in morphology, we suggest that monolayer culture of hBEC represents an experimental model appropriate for many in vitro studies. Moreover, monolayer culture has the advantage that cells can be manipulated for routine tissue culture procedures, such as serial culture and frozen storage. While co-culture with hepatocytes clearly enhanced the growth of hBEC, hBEC had no detectable effect on the growth, survival or function of hepatocytes. This is in agreement with Clement et al. [39], who demonstrated that co-culture of human hepatocytes with autologous human gall bladder epithelial cells had no effect on the survival or albumin production by hepatocytes. Conversely, others have reported increased growth, survival, P450 and albumin production by rat hepatocytes co-cultured with other rat liver `epithelial cells' or non-parenchymal cells [40,41]. These apparently anomalous ®ndings may represent species differences between humans and rodents, but could relate to the relative purity of the co-cultured cells. Semi-puri®ed liver `epithelial cells' or non-parenchymal cells may have included contaminating cell types, including ®broblasts, which in addition to producing mitogens such as HGF, could also secrete a range of other hepatotrophic cytokines capable of improving growth, survival and/or function of hepatocytes. Observations of co-cultured hBEC and hepatocytes [19] suggest that epithelial-derived factors produced by hepatocytes may contribute to the maintenance of morphological polarity of hBEC, while we have found that the mesenchymal-derived factor, hHGF, induces growth at the expense of morphological preservation. Thus, to avoid the different effects of mesenchymal- and epithelial-derived factors, high purity of isolated cell populations becomes of paramount importance, and possible sources of contaminating cells should be minimized. In conclusion, three-dimensional culture of hBEC in collagen gel provides an environment that closely resembles the conditions normally experienced by these cells in vivo. In other studies, the maintenance of three-dimensional interactions between cells, e.g. the use of hepatocyte couplets rather than single cells [42] or intact bile duct units from rodent liver [43,44], has contributed to a better understanding of the functional biology. The three-dimensional model described here allows controlled experimental manipulation of the in vitro environment, while maintaining important three-dimensional interactions, including maintained differentiation between baso-lateral and apical membranes. That secretory activity of hBEC is stimulated in the presence of forskolin suggests that this model will be appropriate to conduct functional studies such as hormonal control of choleresis, and similar studies to those already conducted in rats [15]. One disadvantage of monolayer culture is the inability to access the basal membrane. This membrane is of interest for certain investigations, such as immunological recognition, adhesion and transport studies, e.g. transcytosis of dimeric immunoglobulin A. We suggest that the model described here will

permit studies that require medium-term maintenance of hBEC in culture, while also requiring access to and functional polarization of the basal plasma membrane. Acknowledgements For technical assistance, the authors would like to thank Mr D. Palmer, Department of Clinical Chemistry, Queen Elizabeth Hospital, Birmingham (gGT measurements), Mr A. Keogh, Liver Research Laboratories, University of Birmingham (densitometry scanning) and Mrs L. Tomkins and Mr P. Whittle, Microscopy Services, University of Birmingham (electron microscopy). The authors also thank Professor P. McMaster, Mr J. Buckels and Mr D. Mayer for their help in obtaining normal liver tissue from hepatectomy specimens. This work was funded in part by grants from the Medical Research Council and Endowment Fund of the United Birmingham Hospitals. The support of Telethon Grant, number E 873, and MURST, 1988, number 9806210866, is gratefully acknowledged by M. Strazzabosco. L. Fabris gratefully acknowledges support from Telethon Grant number E-1253. References [1] Daust R. Liver function. In: Brauer RE, editor. American Institute of Biological Sciences, Waverly Press, 1958. p. 3. [2] Strazzabosco M, Joplin R, Zsembery A, Wallace LL, Spirili C, Fabris L, et al. Na 1-dependent and -independent Cl 2/HCO32 exchange mediate cellular HCO3 transport in cultured human intrahepatic bile duct cells. Hepatology 1997;25:976±985. [3] Moorland CM, Fear J, McNab G, Joplin R, Adams DH. Promotion of leucocyte transendothelial cell migration by chemokines from human biliary epithelial cells in vitro. Proc Assoc Am Physicians 1997;109:372±382. [4] Yasoshima M, Kono N, Sugawara H, Katayanagi K, Harada K, Nakanuma Y. Increased expression of interleukin-6 and tumor necrosis factor-alpha in pathologic biliary epithelial cells: in situ and culture study. Lab Invest 1998;78:89±100. [5] Leon MP, Bassendine MF, Gibbs P. Immunogenicity of biliary epithelium; study of adhesive interaction with lymphocytes. Gastroenterology 1997;112:968±977. [6] Leon MP, Bassenedine MF, Wilson JL, Ali S, Thick M, Kirby JA. Immunogenicity of biliary epithelium: investigation of antigen presentation to CD4 1 T-cells. Hepatology 1996;24:561±567. [7] Leon MP, Kirby JA, Gibbs P, Burt A, Bassendine MF. Immunogenicity of biliary epithelial cells ± study of expression of B7 molecules. J Hepatol 1995;22:591±595. [8] Joplin R, Lindsay JG, Johnson GD, Strain AJ, Neuberger JM. Membrane dihydrolipoamide acetyltransferase (E2) on human biliary epithelial cells in primary biliary cirrhosis. Lancet 1992;339:93±94. [9] Joplin R, Wallace LL, Lindsay JG, Palmer JM, Yeaman SJ, Neuberger JM. The human biliary epithelial cell plasma membrane antigen in primary biliary cirrhosis: pyruvate dehydrogenase X? Gastroenterology 1997;113:1727±1733. [10] Demetris AJ, Markus B, Saidman S, Fung JJ, Makowa L, Graner SD, et al. Isolation and primary cultures of human intrahepatic bile ductular epithelium. In Vitro Cell Dev Biol 1998;24:464±470. [11] Joplin R, Strain AJ, Neuberger JM. Immunoisolation and culture of biliary epithelial cells from normal human liver. In Vitro Cell Dev Biol 1989;25:1189±1192.

Y. Ishida et al. / Journal of Hepatology 35 (2001) 2±9 [12] Joplin R, Strain AJ, Neuberger JM. Biliary epithelial cells from the liver of patients with primary biliary cirrhosis: isolation, characterization, and short-term culture. J Pathol 1990;162:255±260. [13] Joplin R, Hishida T, Tsubouchi H, Daikuhara Y, Ayres R, Neuberger JM, et al. Human intrahepatic biliary epithelial cells proliferate in vitro in response to human hepatocyte growth factor. J Clin Invest 1992;90:1284±1289. [14] Bissell DM, Guzelian PS. Phenotypic stability of adult rat hepatocytes in primary monolayer culture. Ann N Y Acad Sci 1980;349:85±98. [15] Vroman B, LaRusso NF. Development and characterization of polarized primary cultures of rat intrahepatic bile duct epithelial cells. Lab Invest 1996(74):303±313. [16] Jacquot J, Spilmont C, Burlet H, Fuchey C, Buisson AC, Tournier JM, et al. Glandular-like morphogenesis and secretory activity of human tracheal gland-cells in a 3-dimensional collagen gel matrix. J Cell Physiol 1994;161:407±418. [17] Wang AZ, Wang JC, Ojakian GK, Nelson WJ. Determinants of apical membrane formation and distribution of multicellular epithelial MDCK cysts. Am J Physiol 1994;267:473±481. [18] Yang L, Faris RA, Hixson DC. Long-term culture and characteristics of normal rat liver bile duct epithelial cells. Gastroenterology 1993;104:840±852. [19] Auth MKH, Okamoto M, Ishida Y, Joplin R, Strain AJ, Markus BH, et al. Co-culture of human hepatocytes and biliary epithelial cells in collagen gel: formation of three-dimensional, liver-like structures in vitro. Langenbecks Arch Chir 1997;114:163±166. [20] Dunn JCY, Yarmush ML, Koebe HG, Tompkins RG. Hepatocyte function and extracellular matrix geometry: long-term culture in a sandwich con®guration. FASEB J 1989;3:174±177. [21] Momburg F, Moldenhauer G, Hammerling GH. Immunohistochemical study of the surface expression of a Mr 34 000 human epitheliumspeci®c glycoprotein in normal and malignant tissues. Cancer Res 1987;47:2882±2891. [22] Strain AJ, Ismail T, Tsubouchi H, Arakaki N, Hishida T, Kitamura N, et al. Native and recombinant human hepatocyte growth factors are highly potent promoters of DNA synthesis in both human and rat hepatocytes. J Clin Invest 1991;87:1853±1857. [23] Joplin R, Lindsay JG, Hubscher SG, Johnson GD, Shaw J, Strain AJ, et al. Distribution of dihydrolipoamide acetyltransferase (E2) in the liver and portal lymph nodes of patients with primary biliary cirrhosis: an immunohistochemical study. Hepatology 1991;14(3):442±447. [24] Rutenberg AM, Kim H, Fischbein JW, Hanker JS, Wasserkrug HL, Seligman AM. Histochemical and ultrastructural demonstration of gamma-glutamyl transpeptidase activity. J Histochem Cytochem 1969;17:517±526. [25] Rothlein R, Czajkowski M, O'Neill MM, Marlin SD, Mainol® E, Merluzzi VJ. Induction of intercellular adhesion molecule 1 on primary and continuous cell lines by pro-in¯ammatory cytokines. J Immunol 1988;141:1665±1669. [26] Hu WN, Blazar BR, Manviel JC, Paradis K, Sharp HL. Phenotypical and functional characterization of intrahepatic bile duct cells: effect of pro-in¯ammatory cytokines. J Lab Clin Med 1996;128:536±544. [27] Hreha G, Jefferson DM, Yu CH, Grubman SA, Alsabeh R, Geller SA, et al. Immortalized intrahepatic mouse biliary epithelial cells: immunologic characterization and immunogenicity. Hepatology 1999;30:358± 371. [28] Zarnegar R, Michalopoulos GK. Puri®cation and partial characterization of human hepatopoeitin A, a polypeptide growth factor for hepatocytes. Cancer Res 1989;49:3314±3320. [29] Ishigawa T, Danda S, Kanetake H, Saitoh Y, Ichihara A, Tomita Y, et al. Hepatocyte growth factor is a potent mitogen for cultured rabbit

[30]

[31]

[32] [33] [34]

[35] [36]

[37]

[38]

[39]

[40] [41] [42]

[43] [44] [45]

9

renal tubular epithelial cells. Biochem Biophys Res Commun 1991;174:831±838. Kan M, Zhang G, Zarnegar R, Michalopoulos G, Myoken Y, McKeehan WL, et al. Hepatocyte growth factor/hepatopoietin A stimulates the growth of rat kidney proximal tubule epithelial cells (RPTE), rat non-parenchymal liver cells, human melanoma cells, mouse keratinocytes and stimulates anchorage independent growth of SV40-transformed RPTE. Biochem Biophys Res Commun 1991;174:331±337. Rubin JS, Chan AML, Bottaro DP, Burgess WH, Taylor WG, Cech AC, et al. A broad-spectrum human lung ®broblast derived mitogen is a variant of hepatocyte growth factor. Proc Natl Acad Sci USA 1991;88:415±419. Johnson M, Koukoulis G, Matsumoto K, Nakamura Tyler A. Hepatocyte growth factor induces proliferation and morphogenesis in nonparenchymal epithelial liver cells. Hepatology 1993;17:1052±1061. Blair JB, Miller MR, Pack D, Barnes R, Teh SJ, Hinton DE. Isolated trout liver cells ± establishing short-term primary cultures exhibiting cell-to-cell interactions. In Vitro Cell Dev Biol 1990;26:237±249. Fabris L, Strazzabosco M, Crosby HA, Ballardini G, Hubscher SG, Kelly DA, et al. Characterization and isolation of ductular cells coexpressing neural cell adhesion molecule and Bcl-2 from primary cholangiopathies and ductal plate malformations. Am J Pathol 2000;156:1599±1612. Milani S, Herbst H, Schuppan D, Niedobitek G, Kim KY, Stein H. Vimentin expression of newly formed rat bile duct epithelial cells in secondary biliary ®brosis. Virchows Arch 1989;415:237±242. Ayres RCS, Neuberger JM, Shaw J, Joplin R, Adams DH. Intercellular adhesion molecule-1 and MHC antigens on human intrahepatic bile duct cells: effect of pro-in¯ammatory cytokines. Gut 1993;34:125±129. Saidman SL, Duquesnoy RJ, Zeevi A, Fung JJ, Starzl TE, Demetris AJ. Recognition of major histocompatibility complex antigens on cultured biliary epithelial cells by alloreactive lymphocytes. Hepatology 1991;13:239±246. Zsembery A, Spirili C, Sitter G, Melzi ML, Colombo C, Stazzabosco M, et al. Ca 21 activated Cl 2 channels can substitute for CFTR in stimulation of anion exchanger in human isolated cholangiocytes Abstract. J Hepatol 1999;30:141. Clement B, Guguen-Guillouzo C, Campion JP. Long-term co-cultures of human adult hepatocytes with rat liver epithelial cells: modulation of albumin secretion and accumulation of extracellular material. Hepatology 1984;4:373±380. Begue JM, Gugen-Guillouzo C, Pasdeloup N. Prolonged maintenance of active cytochrome P-450 in adult rat hepatocytes co-cultured with another liver cell type. Hepatology 1984;4:839±842. Gugen-Guillouzo C, Clement B, Baffet G. Maintenance and reversibility of active albumin secretion by adult rat hepatocytes co-cultured with another liver epithelial cell type. Exp Cell Res 1983;143:47±54. Wilton JC, Matthews GM, Burgoyne RD, Mills CO, Chipman JK, Coleman R. Fluorescent choleretic and cholestatic bile-salts take different paths across the hepatocyte ± transcytosis of glycolithocholate leads to an extensive redistribution of Annexin-II. J Cell Biol 1994;127:401±410. Cho WK, Mennone A, Boyer JL. Intracellular pH regulation in bombesin-stimulated secretion in isolated bile duct units from rat liver. Am J Physiol 1998;38:1028±1036. Gong AY, Larusso NF. Development and initial application of intact bile duct units isolated from normal mouse liver Abstract. Hepatology 1999;30:1081. MartõÂnez-Anso E, Castillo JE, DõÂez J, Medina JF, Prieto J. Immunohistochemical detection of chloride/bicarbonate anion exchangers in human liver. Hepatology 1994;19:1400±1406.