Materials and methods

Liver specimens

All human liver specimens consisted of rest material of liver explants taken for diagnostic/therapeutic purposes. The liver explants were diagnosed for acute hepatitis with submassive liver cell necrosis (AH), post-HCV liver cirrhosis (HCV), and PBC in Ludwig stage IV. The aetiology for the AH group included acute hepatitis B infection as well as drug-induced AH. Resection specimens from normal-appearing liver with no histological abnormalities were derived from liver sections surrounding focal nodular hyperplasia or colon metastasis and used as control samples (H). For routine diagnosis samples were fixed in 6% formalin or in B5 fixative and embedded in paraffin. For microdissection and RNA isolation, samples were embedded in OCT compound (Miles, Kankakee, Illinois, USA), snap-frozen in liquid nitrogen cooled isopentane and stored at −80°C until further use. For assessment of histopathology and fibrosis, the sections were stained with haematoxylin–eosin, PAS diastase, Hall's bilirubin and Sirius red.

Keratin 7 antibody-guided laser capture microdissection

All buffers and solutions were prepared with diethylenepyrocarbonate (DEPC)-treated water (Ambion, Austin, Texas, USA). All used glassware was treated with RNase Zap (Ambion) and washed three times with DEPC-treated water prior to use. Antibody incubations and DAB procedures were performed in the presence of 0.4 U/μl SUPERase RNase Inhibitor (Ambion). Thick frozen sections (10 μm) of human liver (n=4 per group) were cut using RNase-free blades, mounted on precooled (4°C) RNase-free PEN membrane slides (P.A.L.M.; MicroLaser Technologies, Bernried, Germany), immediately placed on dry ice, and stored for up to 1 week at −80°C. Rapid immunohistochemical staining was performed on one slide at a time, to optimise speed of handling. Before laser capture microdissection (LCM), frozen sections were taken from −80°C storage, immediately fixed in ice-cold acetone (−20°C for 5 min) and washed briefly (3–5 s) in phosphate-buffered saline (PBS). Sections were incubated with anti-human K7 (1:20 in PBS for 7 min at room temperature (RT); Dako, Glostrum, Denmark; M7018), briefly washed in PBS, and subsequently incubated with EnVision goat anti-mouse peroxidase-conjugated antibody (undiluted for 7 min at RT; Dako; K4006). The signal was developed with DAB solution (0.06% 3,3′-diaminobenzidine in PBS, 0.003% H₂O₂) for 3 min. Sections were dehydrated in an ethanol series (75–95–100%, 15 s each). The LCM procedure was performed on the P.A.L.M. Microbeam II with an Axiovert-200 microscope (P.A.L.M.). Keratin 7 reactive ductules were dissected (for examples see supplementary file 1), avoiding cholestatic hepatocytes (in PBC) and differentiating intermediate hepatocytes in AH and HCV (for examples see supplementary file 2). The laser-dissected cells were then ejected from the slide with a single defocused laser pulse and catapulted directly into an adhesive cap (P.A.L.M.). Cells were collected within 25 min. Afterwards, 50 μl of lysis buffer (Ambion) was added to the adhesive cap and vortexed upside down. Lysates were collected by centrifugation (14 000 g, 5 min), snap frozen in liquid nitrogen and stored at −80°C until further use.

RNA isolation and amplification

From each liver sample, four LCM samples were pooled for total RNA extraction using the RNAqueous-Micro kit (Ambion). The RNA isolation included an additional DNase treatment (0.1 U/μl) to exclude DNA contamination (Ambion). RNA quality after LCM was determined with an Agilent BioAnalyzer-2100 (Agilent, Palo Alto, California, USA) in combination with RNA 6000 Pico-LabChip. The RNA was amplified with the WT-Ovation Pico RNA Amplification System (NuGEN Technologies, Bemmel, The Netherlands). The WT-Ovation RNA amplification system reverse transcribes RNA to cDNA, which is amplified during the so-called SPIA amplification, a linear isothermal DNA amplification process,32 resulting in end-products of single-strand DNA (ssDNA). The amplified product was purified with DNA Clean&Concentrator-25 kits from Zymo research (Baseclear Lab Products, Leiden, The Netherlands). The amplified products were diluted three times and stored at −20°C until use.

Gene-expression measurements

All qPCR experiments were conducted with a MyiQ Single-Colour Real-Time PCR Detection System (BioRad, Veenendaal, the Netherlands). For PCR array experiments, a RT2 Profiler Custom PCR Array was used to simultaneously examine the mRNA levels of the genes of interest and five reference genes (see supplementary file 3), according to the protocol of the manufacturer (SuperArray Bioscience, Frederick, Maryland, USA). A total volume of 25 μl of PCR mixture, which included 12.5 μl of RT2 Real-Time SYBR Green/Fluorescein PCR master mix from SuperArray Bioscience, 11.5 μl of milliQ, and 1 μl of template cDNA, was loaded in each well of the PCR array. PCR amplification was conducted with an initial 10-min step at 95°C followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Data were analysed using the comparative cycle threshold method with normalisation of the raw data to reference genes including 18S ribosomal RNA (18SrRNA), hypoxanthine phosphoribosyltransferase 1 (HPRT), ribosomal protein L13a (RPL13A), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and β-actin (ACTB).

Statistical analysis of gene expression measurements

For gene-expression measurements, the relative gene expression of each gene-product was used as the basis for all comparisons. The results were assessed for normality using the Kolmogorov–Smirnoff and the Levene tests. The data were not normally distributed; therefore the non-parametric (Kruskal–Wallis) analysis of variance was used for all comparisons. Post-test analysis was performed with a Mann–Whitney U test. Analysis was performed using SPSS software (SPSS Benelux, Gorinchem, The Netherlands). A p value less than 0.05 was considered statistically significant.

Immunohistochemistry

IHC was performed both on serially cut cryostat sections (5 μm) and on serially cut paraffin embedded sections (4 μm). For IHC on snap frozen samples (H (n=5), AH (n=10), HCV (n=10), and PBC (n=10)), sections were fixed in acetone for 10 min. Endogenous peroxidase activity was blocked in Dual Endogenous-Enzyme Blocking Reagent (Dako) for 10 min at RT. Afterwards, sections were rinsed in PBS (pH 7.4) and incubated with primary antibody for 1 h at RT (table 1). Samples were rinsed with PBS for 5 min, then incubated for 30 min at room temperature with EnVision (Dako), developed with AEC (Dako) and counterstained with Mayer's haematoxylin. For IHC on paraffin sections (H (n=5), AH (n=10), HCV (n=10), and PBC (n=10)), endogenous peroxidase activity was blocked in methanolic hydrogen peroxide (2.5%) for 30 min at RT. Antigen retrieval was performed as described in table 1. For heat-induced antigen retrieval sections were incubated in citrate buffer (Dako S2031) for 30 min in a water bath at 96°C and cooled down 10 min at RT. Sections were rinsed in PBS (pH 7.4) and incubated with primary antibody for 1 h at RT or at 4°C overnight (table 1). Samples were rinsed with PBS for 5 min, and incubated for 30 min at room temperature with EnVision (Dako). The signal was developed in DAB solution (0.06% 3,3′-diaminobenzidine in PBS, 0.003% H₂O₂) for 5 min and counterstained with Mayer's haematoxylin. Negative controls included omission of the primary antibody.

Statistical analysis of immunohistochemistry

Evaluation of IHC was performed by calculating the average number of IHC-positive HPCs per field using a 20 times objective counted in three to five non-overlapping fields. IHC determination was performed independently and blindly by two independent investigators (TR and GC). Data were expressed as mean ± standard deviation (SD). Statistics were performed using the Mann–Whitney U test and Spearman's correlation coefficient. A p value less than 0.05 was considered statistically significant.

Immunofluorescence

In order to confirm the gene-expression profiling of Wnt and Notch activation, downstream nuclear transcription factors (β-catenin and Notch-1 intracellular domain (Notch1/NICD) were examined. Immunofluorescence double staining was performed to identify (nuclear) staining in K7 positive cells on 10 μm cryosections, with parallel antibody incubations directed against K7 and β-catenin, or K7 and Notch1/NICD. The nuclei were stained with ToPro3. Slides (AH (n=4), HCV (n=4), and PBC (n=4)) were incubated with mixed primary antibodies overnight at 4°C (table 1), and with mixed secondary antibodies, Alexa 488 goat anti-mouse and Alexa 568 goat anti-rabbit (Invitrogen, Breda, The Netherlands), at RT for 30 min. Counterstain was performed with ToPro-3 (Invitrogen) for 20 min. Rinsing steps were performed in TBS with 0.1% Tween 20, and slides were covered with Aquamount (Vector Laboratories, Burlingame, California, USA). Pictures were obtained by a Leica TCS SP confocal microscope (Leica microsystems, Rijswijk, the Netherlands) with an argon laser at 488 nm, a krypton laser at 568 nm, and a helium/neon laser at 633 nm.

Results

Gene-expression profiling

Gene expression profiles (figures 1–3) were determined from microdissected (activated) liver progenitor cells/ductular reaction (HPCs/DR) from several liver diseases; AH, (n=4), HCV (n=4), and PBC (n=4). Examples of microdissected HPCs/DR see supplementary figure 1. RNA quality measurements (RIN value from 0 to 10) indicated good quality RNA starting material (range 6.8–8.4), and medium quality RNA after microdissection (range 4.0–6.9). Expression of stem cell markers including prominin 1 (CD133), α-fetoprotein (AFP), neural cell adhesion molecule (NCAM), keratin 19 (KRT19), keratin 7 (KRT7), ATP-binding cassette, subfamily G, member 2 (ABCG2/BCRP), leukaemia inhibitory factor V-KIT Hardy–Zuckerman 4 feline sarcoma viral oncogene homolog (KIT/SCFR), Pou domain, class 5, transcription factor 1 (OCT4/POU5F1), CD44 antigen (CD44), oestrogen-related receptor β (ESRRB), homeobox transcription factor NANOG (NANOG), and telomerase reverse transcriptase (TERT), were all present in microdissected HPCs/DR (figure 1). CD133 and AFP were more prominently expressed in AH versus HCV and PBC. NCAM was highly expressed in AH and PBC but lower in HCV. Both KRT19, KRT7 and LIF were highly expressed in HCV although not statistically different towards PBC. OCT4 was more prominently expressed in HCV versus AH and PBC. ABCG2, KIT, CD44, and TERT were similarly expressed in the disease groups. From the ‘stemness’ genes, gene expression of ESRRB and NANOG was induced in PBC compared to AH and HCV. The most prominent expressed mesenchymal marker in all three groups was glial fibrillary acidic protein (GFAP), while desmin (DES), a marker of myofibroblasts, was more prominently expressed in PBC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Gene-expression profile of stem-cell markers. Bar graphs indicate relative expression of mRNA levels in acute hepatitis with submassive liver cell necrosis (AH, n=4), post-HCV liver cirrhosis (HCV, n=4), and primary biliary cirrhosis in Ludwig stage IV (PBC, n=4). Different letters above bars indicate statistical significance (p<0.05). ABCG, ATP-binding cassette, subfamily G, member 2 (BCRP); AFP, α-fetoprotein; DES, desmin; ESRRB, oestrogen-related receptor beta; GFAP, glial fibrillary acidic protein; KRT, keratin; LIF, leukaemia inhibitory factor; KIT, V-KIT Hardy–Zuckerman 4 feline sarcoma viral oncogene homologue (SCFR); NANOG, homeobox transcription factor NANOG; NCAM, neural cell adhesion molecule; OCT, Pou domain, class 5, transcription factor 1 (POU5F1); TERT, telomerase reverse transcriptase.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Gene-expression profile of Wnt signalling pathway. Bar graphs indicate relative expression of mRNA levels in acute hepatitis with submassive liver cell necrosis (AH, n=4), post-hepatitis C virus liver cirrhosis (HCV, n=4), and primary biliary cirrhosis in Ludwig stage IV (PBC, n=4). Different letters above bars indicate statistical significance (p<0.05). APC, adenomatous polyposis coli; AXIN2, axis inhibitor 2; FZD1, frizzled 1; GSK3B, glycogen synthase kinase 3-beta; LEF1, lymphoid enhancer-binding factor 1; WNT, wingless-type MMTV integration site family.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Gene-expression profile of the Notch signalling pathway. Bar graphs indicate relative expression of mRNA levels in acute hepatitis with submassive liver cell necrosis (AH, n=4), post-hepatitis C virus liver cirrhosis (HCV, n=4), and primary biliary cirrhosis in Ludwig stage IV (PBC, n=4). Different letters above bars indicate statistical significance (p<0.05). DLL, delta-like; JAG, Jagged; NOTCH, Notch, Drosophila, homologue of;

From the Wnt signalling pathway (figure 2), the ligands wingless-type MMTV integration site family (WNT)3A and WNT5 were most prominently expressed in AH, HCV and PBC. Wnt receptor frizzled 1 (FZD1) was more prominently expressed in HCV although not statistically different towards PBC. WNT1, WNT4 and axis inhibitor 2 (AXIN2) were all equally expressed in AH, HCV and PBC. Glycogen synthase kinase 3-β (GSK3B) expression was elevated in HCV although not statistically different form AH. The lowest expression of adenomatous polyposis coli (APC) (necessary for downregulation of β-catenin) was found in AH while it increased significantly in HCV and was highly expressed in PBC. Lymphoid enhancer-binding factor 1 (LEF1), a transcription factor from the T cell-specific transcription factor/lymphoid enhancer factor-1 (TCF/LEF1) family, which is activated by intracellular protein β-catenin, was significantly expressed in all diseases with the highest expression in PBC.

From the Notch signalling pathway (figure 3) δ-like (DLL) 1 was highly expressed in AH whereas Jagged-1 (JAG1), from the serrate ligands, was upregulated in PBC. DLL3, DLL4 and Jagged-2 (JAG2) were equally expressed in AH, HCV and PBC. Notch, drosophila, homolog of (NOTCH) 1 expression was highest in HCV, while NOTCH4 expression in HCV was lower than PBC. No differences in NOTCH2 and NOTCH3 expression was found in AH, HCV and PBC. The most prominent receptors were NOTCH1 and NOTCH3.

Immunohistochemistry

To confirm gene-expression profiles and to compare to histologically healthy liver samples, IHC for K7, NCAM, prominin-1 (CD133), β-catenin, LEF1 (transcription factor of an active Wnt signalling pathway), and Jagged-1 (JAG1) was performed on a larger (n=30) series of patients. Examples of immunohistochemical findings are provided in figure 4. The average amounts of positive HPCs per microscopic field are provided in tables 2 and 3.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Immunohistochemical staining. Examples of K7 (A–D), CD133 (E–H), NCAM (I–L), JAG1 (M–Q), β-catenin (P–T) and LEF1 (U–X) stainings in histologically healthy livers (H), acute hepatitis with submassive liver cell necrosis (AH, n=4), post-hepatitis C virus liver cirrhosis (HCV, n=4), and primary biliary cirrhosis in Ludwig stage IV (PBC, n=4). Original magnification: ×20 (A–P, U–X) or ×40 (Q–T). Arrows indicate progenitor cells and arrowheads interlobular bile ducts. CD133, prominin 1; JAG1, Jagged-1; K7, keratin 7; LEF1, lymphoid enhancer-binding factor 1; NCAM, neural cell adhesion molecule.

In histologically healthy liver, small amounts of K7 positive cells are seen in the periportal region (10.2±1.3, figure 4A). Compared to healthy livers, the number of K7 positive cells is diffusely increased in AH (72.9±8.2, figure 4B) and HCV (41.3±5.8, figure 4C), and is increased in the fibrous septa and (septal) interface in PBC (45.3±6.1, figure 4D). In histologically healthy liver, scarce numbers of CD133 positive HPCs with an apical staining pattern are located in the canals of Hering (1.2±0.5, figure 4E). Besides these HPCs and some occasional CD133 positive cholangiocytes, no other cell types were positive. In AH, the highest amount of CD133 positive HPCs is found in a diffuse pattern (34.0±7.0, figure 4F). CD133 positive HPCs are mainly located into the fibrous septa in HCV (7.3±1.8, figure 4G) and PBC (5.9±1.9 figure 4H). In histologically healthy liver, numerous NCAM-positive lobular hepatic stellate cells and NCAM-positive nerves and nerve endings are seen, while positive NCAM staining of interlobular bile ducts is rare. Some NCAM-positive HPCs were found periportally (2.3±1.1, figure 4I). A high number of NCAM-positive HPCs were diffusely found in AH (64.4±8.9, figure 4J). NCAM-positive HPCs were located in the fibrous septa and interface in HCV (18.1±4.2, figure 4K) and PBC (30.9±5.3, figure 4L). In histologically healthy liver, rare JAG1-positive HPCs, some JAG1-positive bile ducts and no JAG1-positive hepatocytes or endothelial cells are seen (figure 4M). Numerous JAG1-positive HPCs are present in AH (35.6±8.6, figure 4N), HCV (16.6±9.5, figure 4O) and PBC (31.6±6.1, figure 4P). In histologically healthy liver, β-catenin is mostly expressed on the cell membrane of HPCs, cholangiocytes and hepatocytes (figure 4Q). In contrast, β-catenin is expressed in the cytoplasm and nuclei in HPCs in AH (22.0±7.6, figure 4R), HCV (9.6±3.6, figure 4S), and PBC (20.2±7.3, figure 4T). In histologically healthy liver, a moderate diffuse staining of LEF1 in hepatocytes was observed (figure 4U). LEF1 is highly positive in the nuclei of HPCs in AH (35.2±7.1, figure 4V) and in PBC (41.0±11.0, figure 4X), while a lower expression is found in HCV (13.0±4.7, figure 4W).

Immunofluorescence

To confirm a possible activation of Wnt and Notch pathways in HPCs/DR, immunofluorescence double staining was performed by using antibodies directed against K7 and β-catenin (figure 5A) or K7 and Notch1/NICD (figure 5B). In all diseases, the most prominent β-catenin staining was found on parenchymal cells and confirmed the immunohistochemical staining. Double staining of K7 and β-catenin was largely found in AH and to a lesser extent in HCV and PBC. The nuclear presence of β-catenin was mostly in AH (figure 5A). In all diseases, Notch1/NICD staining showed a diffuse cytoplasmic staining of hepatocytes. Some of the bile ducts and some portal fibroblasts stained positive (cytoplasmic). Other portal structures and the central vein were negative. Double staining of K7 and Notch/NICD as well as nuclear localisation of Notch/NICD in HPCs was predominantly found in PBC, whereas little to no nuclear staining of HPCs was found in HCV or AH (figure 5B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Immunofluorescence staining. Examples of immunofluorescence double staining with (A) β-catenin (red) and K7 (green) and (B) Notch1 intracellular domain (NICD, red) and K7 (green) in acute hepatitis with submassive liver cell necrosis (AH, n=4), post-hepatitis C virus liver cirrhosis (HCV, n=4), and primary biliary cirrhosis in Ludwig stage IV (PBC, n=4). Counter stain Topro-3 (blue). Original magnification: 4×40 (insets ×80). Arrows indicate examples of nuclear staining in K7 positive cells.

Discussion

HPCs represent a reserve cell compartment that is activated when the mature epithelial cells of the liver (hepatocytes and cholangiocytes) are damaged or inhibited in their replication.33 In the human liver, HPCs are activated after acute and chronic damage.34 Different subpopulations of HPCs could be described considering their phenotype and their differentiation depending on the type and severity of the disease.35 In the present study, we have used microdissection and gene-expression profiling to study the proliferation and differentiation of HPCs in three different liver diseases: an acute disease with prevalent hepatocellular damage (AH), chronic hepatocellular disease (HCV) and primary biliary disease (PBC). Results of this study led to the proposed model of HPC activation and differentiation as summarised in figure 6.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Proposed model for hepatic progenitor cell activation/differentiation. CD133, prominin 1; HPC, hepatic progenitor cell; K7, keratin 7; NCAM, neural cell adhesion molecule. (A) Schematic figure of HPC activation and differentiation. CD133 loss is a key feature of HPC differentiation. the Wnt pathway activation plays a key role for stem cell niche expansion. the presence of Notch signal drives the hpcs fate choice towards cholangiocytic lineage. (B) Different expression of hpcs markers and Wnt/Notch pathways in the liver diseases under study.

HPCs represent a dynamic cellular compartment and the severity and duration of injury play an important role in HPC proliferation and differentiation. The expression of CD133 is usually found in uncommitted progenitor cells.36 In addition, CD133 correlates with AFP, which has been previously confirmed in hepatocellular carcinomas with CD133 expression.37 In AH, HPCs proliferate in large numbers and diffusely express NCAM, AFP and CD133 suggesting that these HPCs represent a subtype of HPCs during this acute activation after submassive hepatocyte necrosis. In contrast, HPC activation in HCV is less pronounced as indicated by the decreased number of K7-positive HPCs (table 2). In HCV few CD133-positive HPCs were found suggesting a more differentiated (committed towards hepatocytic lineage) state of these cells during this chronic lower degree of hepatocyte damage. In PBC, the HPC compartment has a prominent activation as indicated by the number of K7-positive HPCs. HPCs in PBC have a pronounced NCAM expression; however, fewer CD133-positive HPCs are present indicating that the HPCs are in a more differentiated state (committed towards biliary lineage) compared to AH.38 CD133-positive cells may therefore identify a more restricted population of progenitor cells, which are maximally activated after submassive acute necrosis of hepatocytes. The presence of CD133-positive cells raises the question of whether these cells derive from haematopoietic stem cells or represent the resident HPC compartment.12 39 CD133-positive HPCs have been detected in conditions (eg, healthy liver) in which a bone marrow contribution is absent or minimal suggesting that these cells should represent resident HPCs.40 41 Recently, CD133-positive hepatic stellate cells have been described as progenitor cells in rats.42 Although the possibility of a mesenchymal cell type with CD133 positivity cannot be excluded, the cells in healthy liver appear epithelial in nature. Taken together, CD133 is a marker for activated progenitors in acute phases of hepatocytic liver disease and is correlated with an increased AFP expression.

In AH, activated HPCs predominantly proliferate rather than differentiate.17 43 HPC activation plays a key role in the prognosis of this pathology as indicated previously.44 Although HPCs massively proliferate in AH, the differentiation towards hepatocytes takes place only after 1 week from the initial liver injury. This delay in differentiation in turn affects the outcome of the patients with AH. The Wnt pathway, which has a key role in stimulating stem cell proliferation,45 is more pronounced in AH compared to HCV or PBC (figures 2, 4 and 5). The important role of the Wnt pathway in HPC activation/proliferation was described before in several in vitro assays as well as in murine HPC proliferation models.28 30 46 47 The similar activation of Wnt in AH compared to these toxin-induced murine models underlines the importance of this pathway in human liver diseases with an acute onset.

Terminal differentiation of HPCs is characterised by the loss of NCAM.48 In PBC the majority of HPCs are NCAM positive indicating their decreased capacity for terminal differentiation into mature cholangiocytes (characterised by the loss of NCAM expression), although some NCAM negative HPCs remain. As a result, the prominent HPC activation is not sufficient to restore cholangiocyte loss. In addition, the observation that ‘stemness’ genes ESRRB and NANOG were up regulated in PBC confirm a lack of terminal differentiation.

In haematopoietic, intestinal and neuronal stem cell niche, the modulation of Notch activity is fundamental in influencing the fate of progenitor cells.45 The role of the Notch pathway in bile-duct development first became apparent in Alagille syndrome,49 where mutations in the Jagged-1 gene (JAG1) were identified as the cause of Alagille syndrome causing bile-duct paucity. In general, Notch2 is responsible for the morphogenesis of forming bile-duct structures.20 23 Notch 1 and 3 play an important role in biliary differentiation and maintenance in murine models.31 In the present study lower JAG1 encoding mRNA in hepatocytic diseases (HCV and AH) was measured. Although protein expression of Jagged-1 did not correlate with the gene expression in AH (where it was still present on the HPCs), the lack of nuclear Notch staining in HCV (and increased nuclear staining in PBC) suggests a probable inhibition of this pathway in hepatocytic diseases. The increased values of JAG1 and nuclear Notch1/NICD staining in PBC could be linked with fate choice towards cholangiocytes. This is in agreement with the observation that immature and mature cholangiocytes express members of the Notch pathway in vivo.50 In rodent models Notch was found to play an important role in biliary differentiation of HPCs in a model of biliary ductopenia,20 23 and after partial hepatectomy.51 52 Our data suggest that Notch plays a similar role in biliary differentiation of human HPCs.

In most conditions the Notch signal requires physical contact between cells.53 54 The observation that desmin positive myofibroblasts are mostly present surrounding HPCs during PBC indicates that these cells might be responsible for the (Notch) cell–cell communication between progenitor cells and mesenchymal cells regulating their bipotential cell fate which remains to be determined by functional assays.

Other pathways which were found to be of interest in this study included BMI1, a transcriptional receptor belonging to the polycomb group gene family,55 which was present in all groups with a slight (non-significant) increase in the PBC group (data not shown). In the liver, BMI1 has been associated with ductopenia and an increased malignancy of HCC.56 57 Due to the involvement of stem cell self-renewal of BMI1, this could indicate the increased self-renewal capacity of HPCs during liver disease.58 In addition, from the hedgehog signalling pathway both Desert hedgehog as well as Sonic hedgehog were both expressed by HPCs (data not show). Recently Sonic hedgehog has been shown to be involved in epithelial-to-mesenchymal transition of ductular cells in non-alcoholic fatty liver disease.59 If epithelial-to-mesenchymal transition does occur in HPCs during AH, HCV, and PBC remains to be proven.

Controlled manipulation of the HPC niche holds great promise for therapeutic intervention in currently untreatable liver diseases.60–62 Although some of these treatments (such as γ-secretase inhibitors) mostly rely on Wnt or Notch inhibition in neoplastic diseases,63 these treatments might be used to stimulate a patient's own progenitor cell compartment.

In conclusion, our gene expression analysis and immunohistochemical study on human liver diseases indicate that: (1) CD133 represents a surface marker for a restricted population of expanding HPCs associated with high AFP expression; (2) activation of the Wnt pathway plays a significant role in HPC expansion; and (3) the Notch pathway is involved in the fate choice of HPCs towards the cholangiocytic lineage. Functional assays should confirm these findings. These results hold great promise for development of future therapeutic strategies.