Combined hepatocellular cholangiocarcinoma originating from hepatic progenitor cells: immunohistochemical and double-fluorescence immunostaining evidence


X-P Chen, Hepatic Surgery Centre, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430030, China.


Aims:  Combined hepatocellular cholangiocarcinoma (CHC) is a rare form of primary liver cancer, showing a mixture of hepatocellular and biliary features. Data suggest that most CHC arise from hepatic progenitor cells (HPCs). The aim was to investigate the origin of CHC.

Methods and results:  Twelve cases of CHC were studied by immunohistochemistry for hepatocytic (hepPar1, α-fetoprotein), cholangiocytic cytokeratin [(CK) 7, CK19], hepatic progenitor cell (OV-6), haematopoietic stem cell (c-kit, CD34), as well as CD45 and chromogranin-A markers. The combination of double-fluorescence immunostaining consisted of HepPar1 with CK19, and c-kit with OV-6. All 12 cases demonstrated more or less transitional areas, with strands/trabeculae of small, uniform, oval-shaped cells including scant cytoplasm and hyperchromatic nuclei embedded within a thick, desmoplastic stroma; however, two cases were found to consist entirely of such transitional areas. Simultaneous co-expression of hepPar1 and CK7, or CK19, was demonstrated in 10/12 (83.3%) cases of CHC. c-kit expression was noted in 10/12 (83.3%) cases, of which 7/10 (70%) showed co-expression of OV-6.

Conclusions:  The results suggest that CHC are of HPC origin, supporting the concept that human hepatocarcinogenesis may originate from the transformation of HPCs.


cholangiocellular carcinoma


combined hepatocellular cholangiocarcinoma




haematoxylin and eosin


hepatocellular carcinoma


The two major primary liver cancers in adults are hepatocellular carcinoma (HCC) and cholangiocellular carcinoma (CC). A well-documented, although rare, subset of hepatic malignancies, namely combined hepatocellular cholangiocarcinoma (CHC), was initially described by Allen and Lisa1 in 1949 and includes the composition and characteristics of HCC and CC coexisting in the same tumour or liver. According to the World Health Organization classification, the histopathological definition of CHC, as well as its diagnosis, is based on criteria that require unequivocal elements of both HCC and CC.2 However, some CHC tumours exhibit separable areas between HCC and CC and many show intermediate features in most or all of the lesions, which makes classification based on these criteria challenging. Although Allen and Lisa have classified patients with CHC into three subtypes, Goodman et al.,3 Taguchi et al.4 and Libbrecht et al.5 have identified different criteria for the pathomorphological diagnosis of CHC, and classified this disease into three categories, respectively. The classification of CHC is usually based on three categories: type I, collision tumours and double cancers, where the HCC and CC components are either completely separated or sharply demarcated; type II, tumours with contiguous, independent masses of HCC and CC, separated by an intervening area of transition; and type III, tumours consisting entirely of transitional areas, containing only very limited, ambiguous hepatocellular and cholangiocellular components. In the present study, CHC, types II and III, were included to enable a simple and strict interpretation of the results.

It has long been controversial whether hepatic malignancies arise from stem cells that undergo a malignant transformation or from the de-differentiation of neoplastically transformed mature hepatocytes. In animal models of hepatocarcinogenesis, the proliferation of small, undifferentiated, oval-shaped cells with scanty cytoplasm has been detected, which appear around portal tracts prior to hepatocyte regeneration or cholangiocyte proliferation. These cells are referred to as oval cells and are thought to be facultative bipotential progenitor cells. Oval cells have also been linked to subsequent development of hepatic malignancies in these animal models.6,7 Recent studies have confirmed the existence of human cell populations with similar localization, morphology, ultrastructure, immunohistochemical phenotype and functional aspects, and have termed them ‘hepatic progenitor cells’ (HPCs).8 These cells have been identified in several human liver diseases, including chronic viral hepatitis, cirrhosis, focal nodular hyperplasia and hepatocellular adenoma.9–12 Evidence indicates that HPCs possess the potential to differentiate into either hepatocytes or cholangiocytes, which suggests that HCCs may originate from HPCs.13 The fact that CHCs contain both hepatocytic and cholangiocytic features is also important evidence to support the hypothesis that hepatic malignancy could be stem cell- or progeny-derived. To investigate the possible function of bipotent HPCs in human CHC development, immunohistochemistry and double-fluorescence immunostaining were performed on paraffin-embedded tissues of 12 consecutively resected CHCs using the following markers: hepatocytic (hepPar1, α-fetoprotein),14,15 cholangiocytic cytokeratin [(CK) 7, CK19],16 hepatic progenitor cells (OV-6),17 haematopoietic stem cells (CD34, c-kit),18 leucocyte common antigen (CD45)19 and the neuroendocrine marker (chromogranin-A)20 to better understand the origin of CHC. CD45 allowed the separation of mast cells from c-kit+ progenitor cells and chromogranin-A positive progenitor cells.

Materials and methods

Patients and tissue specimens

This study was based on 12 consecutively resected CHCs at the Hepatic Surgery Centre of Tongji Hospital (Wuhan, China) between January 1996 and December 2005. Tissue samples were collected following a protocol approved by the Institutional Review Board for Human Research of Tongji Hospital. The clinical data, including serological data and background liver disease, were obtained from hospital records for each case (see Table 1).

Table 1.   Clinical data pertaining to the CHC patients
CaseAge (years)/sexAetiologyLiver fibrosisSerum AFP (μg/l)Tumour locationSize (mm)Operative procedure
  1. AFP, α-fetoprotein; HBV, hepatitis B virus.

 150/MHBVCirrhosis4.32IV 30 × 40Lobectomy
 275/M2.65II, III 60 × 90Lobectomy
 347/F3.58V100 × 120Non-anatomical resection
 442/MHBVCirrhosis2.81VI 65 × 67Non-anatomical resection
 543/FHBVCirrhosis93.07II 30 × 70Segmentectomy
 642/MHBVCirrhosis2.54IV, V110 × 120Non-anatomical resection
 748/F1.81II 60 × 90Lobectomy
 848/M5.85II, III 40 × 45Resection of left half
 942/MHBVCirrhosis1.9IV, V 67 × 73Non-anatomical resection
1050/MHBVCirrhosis4.54V, VI 45 × 60Non-anatomical resection
1145/F3.21II, III 50 × 75Lobectomy
1244/MHBVCirrhosis2.58IV, V 70 × 80Non-anatomical resection

The 12 patients consisted of eight men and four women with a mean age of 48 years (range 42–75 years). Seven (58.3%) of the 12 patients were seropositive for hepatitis B virus surface antigen, and none was seropositive for anti-hepatitis C virus antibody. Seven patients (58.3%) had chronic hepatitis and cirrhosis.

Five cases of surgical specimens of typical HCC, as well as three cases of typical intrahepatic CC confirmed by an experienced pathologist, were selected for comparison of immunohistochemical and double-fluorescence immunostaining studies.

Biopsy specimens were formalin-fixed and paraffin-embedded. Routine haematoxylin and eosin (H&E) staining was performed on 5-μm thick tissue sections. Two experienced pathologists evaluated all histological specimens, according to the histological features of CHC, HCC and CC.


Tissue slices (5 μm) were sectioned from representative paraffin blocks of CHC, subsequently deparaffinized in xylene and rehydrated through graded alcohols. Endogenous peroxidases were inactivated using 3% H2O2 in methanol for 20 min. Sections underwent antigen retrieval and were treated in 10 mm Tris buffer, 1 mm ethylenediaminetetraacetic acid (pH 9.0) for 20 min in a microwave oven at 98.5°C. Sections were then incubated with primary antibody for 60 min at room temperature and subsequently incubated with goat antimouse EnVision (Dako Corp., Carpinteria, CA, USA) for 30 min at room temperature. Sections were then washed three times with phosphate-buffered saline (pH 7.2) for 2 min. The reaction product was developed with 3,3′-diaminobenzidine and counterstained with haematoxylin.

Immunohistochemistry was performed on all tissues with antibodies to a panel of antigens, including HepPar1 (Dako), α-fetoprotein (Dako), CK7 (Zymed Corp., Carlsbad, CA, USA), CK19 (Dako), OV-6 (R&D Corp., MN, USA), c-kit (Santa Cruz Corp., Santa Cruz, CA, USA), CD34 (Zymed), CD45 (Dako) and chromogranin-A (Dako).

Immunoreactivity was semiquantitatively evaluated with grading from 0 to 4+ as follows: 0, no staining; 1+, 5% positive cells; 2+, 6–25% positive cells; 3+, 26–50% positive cells; and 4+, >50% positive cells. For the purpose of statistical evaluation, tumours with a grade of 2+ or more were deemed positive for antigen expression.

Double-fluorescence immunostaining

For antigen co-localization studies, double-fluorescence immunostaining was performed with a sequential fluorescent method. Deparaffinization, rehydration and heat-induced epitope retrieval of CHC tissue sections was the same as previously described for immunohistochemistry. The primary antibody combinations consisted of HepPar1 with CK19, and c-kit with OV-6. Secondary antibodies conjugated with fluorescent conjugates cyanine-3 or fluorescein isothiocyanate were used. Double-labelling was performed using both antibodies on the same section. Incubation of primary antibodies and secondary antibodies was performed using the Vector M.O.MTM Fluorescein Immunodetection kit (Vector Laboratories, Peterborough, UK) and the Avidin/Biotin blocking kit (Vector Laboratories). As a control, the primary antibodies were omitted and results were consistently negative. Immunofluorescence was observed with the confocal microscope Nikon Digital ECLIPSE C1 (Nikon Corp., Tokyo, Japan). An argon laser at 488 nm was used in combination with a helium neon laser at 543 nm to excite the green and red fluorochromes simultaneously. Emitted fluorescence was detected with a 505–530-nm bandpass filter for the green signal, a 560-nm longpass filter for the red signal and 633-nm filter for the blue signal.


Histopathological features

All CHCs that included some transitional or intermediate areas showed strands or trabeculae of small, uniform cells with scant cytoplasm and hyperchromatic nuclei. In other transitional areas, the tumours were arranged in strands with vague gland-like structures against a background of broad desmoplastic stroma, resulting in an ‘antler-like’ appearance, as well as small nests of gland-forming adenocarcinoma cells separated by connective tissue.

The transitional type of CHC comprised two cases (2/12, 16.6%) and displayed solid nests of hepatoid cells with focal glandular architecture separated by broad desmoplastic stroma or columnar-shaped tumour cells arranged in a trabecular or adenoid pattern. In addition, the CHCs presented solid nests of small round-to-oval tumour cells with scant eosinophilic cytoplasm (Figure 1).

Figure 1.

 Histological features and immunohistochemical profiles of combined hepatocellular cholangiocarcinoma (CHC). The solid nests of hepatoid cells and focal glandular architecture (A). The intermediate phenotype of CHC is arranged in strands, with vague gland-like structures against the background of broad desmoplastic stroma, resulting in an ‘antler-like’ appearance, as well as small nests of gland-forming adenocarcinoma cells separated by connective tissues (B, C). The tumour cells of transitional areas show strands or trabeculae of small, uniform cells with scant cytoplasm and hyperchromatic nuclei (D).

Five cases of typical HCC demonstrated tumour cells with abundant eosinophilic cytoplasm, large vesicular nuclei and prominent nucleoli, which were arranged in a trabecular pattern and without significant desmoplasia. In contrast, the three cases of typical CC were tumours with significant desmoplasia and prominent gland formation, lined by cuboidal cells.

Immunohistochemical profiles

All 12 CHCs showed immunoreactivity for HepPar1, CK7 and CK19. HepPar1 was predominantly expressed in the hepatocytic areas and CK7 and CK19 mainly in the glandular areas. The tumour cells of the transitional areas showed simultaneous, cytoplasmic expression of HepPar1 and CK19 in 10 (10/12, 83.3%) cases. All tumour cells from transitional areas were immunoreactive for OV-6. Only two cases of CHC were α-fetoprotein-positive (2/12, 16.6%).

Ten (10/12, 83.3%) cases expressed c-kit, which was noted mainly in small cells within the transitional zone that showed intermediate features; of these, seven (7/10, 70%) showed simultaneous expression of hepatocytic or cholangiocytic markers (Figure 2). CD34+ sinusoidal endothelial cells were present in the centre and periphery of tumour nodules; however, tumour cells positive for CD34 were not found.

Figure 2.

 Immunohistochemical profiles of combined hepatocellular cholangiocarcinoma. The tumour cells show immunoreactivity for HepPar1 (A), α-fetoprotein (B), cytokeratin (CK) 7 (C), CK19 (D), CD34 (E), c-kit (F), OV-6 (G) and CD45 (H).

No tumour cells in CHC showed immunoreactivity for chromogranin A. Controls lacked the primary antibody and displayed no specific immunoreactivity.

The immunoreactivity of control HCC was confirmed for HepPar1. CK7+ and CK19+ tumour cells were found in 4/5 cases of HCC, and 1/5 cases of HCC showed immunoreactivity for OV-6. Control CCs did not present reactivity with HepPar1 antibody. In all the typical cases of HCC and CC, there was no c-kit expression detected.

Double-fluorescence immunostaining profiles

The tumour cells in the intermediate areas showed simultaneous cytoplasmic co-expression of HepPar1 and CK19 in eight of the 12 cases (66.6%) using confocal double-fluorescence immunostaining. Co-expression of c-kit with OV-6 in the same tumour cells in the intermediate areas was seen in seven out of 12 cases (58.3%; Figures 3 and 4).

Figure 3.

 Double-immunofluorescence staining for HepPar1 and cytokeratin (CK) 19 in combined hepatocellular cholangiocarcinoma. The tumour cells show reactivity for HepPar1 (fluorescein isothiocyanate, green; A), CK19 (cyanine-3, red; B) and co-localization for HepPar1 and CK19 (yellow; C).

Figure 4.

 Double-immunofluorescence staining for OV-6 and c-kit in combined hepatocellular cholangiocarcinoma. The tumour cells show reactivity for OV-6 (cyanine-3, red; A), c-kit (fluorescein isothiocyanate, green; B) and co-localization for OV-6 and c-kit (yellow; C).

For the results of immunohistochemistry and the double-immunofluorescence studies, see Table 2.

Table 2.   Immunohistochemistry and double-immunofluorescence profiles
 Hep Par1AFPCK7CK19OV-6CD34c-kitHep Par1/CK19c-kit/ OV-6
  1. AFP, α-fetoprotein; CK, cytokeratin; CHC, combined hepatocellular cholangiocarcinoma; HCC, hepatocellular carcinoma; CC, cholangiocellular carcinoma.

CHC (n = 12)12212121201087
HCC (n = 5)524410100
CC (n = 3)003330000


CHC is a rare form of primary liver cancer showing features of both hepatocytic and biliary epithelial differentiation. The incidence of CHC among primary liver cancers is about 1.0–4.7%.21,22 Liu et al.23 have reported an incidence of 2.0% in Hong Kong. The Liver Cancer Study Group of Japan has shown that the CHC form of cancer constitutes 1.2% of surgical cases and 1.6% of autopsy cases.24 In our Hepatic Surgery Centre, the incidence of CHC is 1.05% among resected primary liver cancers.25

In this study of 12 cases of this particular tumour, two histological forms were encountered. Ten cases were type II, or transitional tumours, in which there were areas of intermediate differentiation and an identifiable transition between HCC and CC. There were two cases of type III tumours, or an intermediate (hepatocyte-cholangiocyte) phenotype, which consisted entirely of transitional areas.

The HepPar1 marker was shown to stain >90% of HCCs; however, the sensitivity decreased with poorer differentiation and the intermediate carcinomas in this study were, in fact, tumours that were too ‘primitive’ to be classified as either HCC or CC.26 Immunohistochemistry for intracellular antigens showed that α-fetoprotein was a specific, but poorly sensitive, marker of hepatocellular differentiation in primary liver cancers, and was positive in only 25–50% of HCCs.

CK7 and CK19 are good markers of biliary epithelial differentiation and are found in 90% of cholangiocarcinomas. Our unpublished data show that CK7 immunoreactivity was detected in 21/27 cases of HCC. HCC that expresses progenitor cell/ductular markers, such as CK7 and CK19, presents a more aggressive clinical course and more frequent and rapid recurrence of disease after surgical treatment.27,28 Studies have shown that a cut-off of 5% CK19+ cells is sufficient for influencing the outcome of the HCC patient.29 There are no specific markers that are absolutely indicative of liver stem cells, whereas OV-6 has proven to be promising in some studies.20

The proto-oncogene, c-kit, encodes a transmembrane tyrosine kinase receptor (CD117), which is structurally similar to the receptors for platelet-derived growth factor and colony-stimulating factor-1. Canals of Hering and proliferating progenitor cells in diseased human liver have been reported to express c-kit. The significance of c-kit expression in hepatic progenitor cells is controversial. A proportion of hepatic progenitor cells have been considered by some authors to originate from bone marrow-derived haematopoietic stem cells.30–32 Thorgeirsson et al.33 have recently stated that haematopoietic stem cells contribute little to hepatocyte formation under either physiological or pathological conditions, although they may provide cytokines and growth factors that promote hepatocyte function through paracrine mechanisms. On the other hand, the selective tyrosine kinase inhibitor, STI 571 (Gleevec), which inhibits c-kit receptor tyrosine kinase activity, has received much attention recently for its positive chemotherapeutic effect against chronic myelogenous leukaemia, gastrointestinal stromal tumours and small cell lung carcinoma. Expression of c-kit has also been considered to have important prognostic implications.34,35 In this study, it was found that a significant proportion of CHC tumour cells, characterized by features more primitive than those of typical HCC or CC, showed c-kit expression by immunohistochemistry. This may indicate a novel role for STI 571 in the treatment of those CHCs that present primitive features and c-kit expression.36 Although CD34 and c-kit have been recognized as markers of bipotential progenitor cells in the liver, CD34+ tumour cells were not detected in the present study. The c-kit+ cells did not co-express CD34 in serial sections of tumour tissue. Crosby et al.18 have also suggested that c-kit+ and CD34+ cells are two distinct populations, since they did not co-localize the two markers in serial sections using confocal double-fluorescence immunostaining.

It has been suggested that chromogranin-A may be a marker of hepatic progenitor cells.37 In our study, this marker did not react with any tumour cells, which may relate to formalin fixation, rather than the differentiation status of the tumour cells, as an earlier study using this marker succeeded in detecting immunoreactivity in frozen tissue.38

Another characteristic of the intermediate phenotype of CHC is the broad desmoplastic stroma, which encases the strands and trabeculae of tumour cells and is also an important characteristic of typical CC. This may have important clinical implications, since transhepatic arterial chemoembolization, a widely-used method for the treatment of HCC, may not be as effective in the treatment of intermediate CHC that displays extensive desmoplasia.

Hepatic progenitor cells can be identified by co-expression of hepatocytic and cholangiocytic lineage markers (HepPar1 and CK19) as well as haematopoietic stem cell markers, such as c-kit and CD34.39,40 Double-fluorescence immunostaining for stem cell and lineage markers was used to identify putative hepatic stem or progenitor cells. The immunohistochemical and double-fluorescence immunostaining phenotypes of CHC and hepatic progenitor cells show overlap, indicating that hepatic progenitor cells may play a role in human CHC development. These findings are in accordance with the established function of oval cells in several rodent models of hepatocarcinogenesis.41 It has been demonstrated in vitro that a single clonally expanded cell from a CHC cell line can give rise to both glandular and hepatocytic elements.42 Recent research has shown that in the cirrhotic stage of most chronic liver diseases, hepatocytes become senescent due to telomere reduction.43,44 This provides additional evidence that at least some of the hepatic carcinomas originate from a progenitor cell.

In conclusion, it is proposed that CHCs have distinct clinicopathological features, which are morphologically and phenotypically intermediate between HCC and CC. This supports the hypothesis that CHC originates from hepatic stem or progenitor cells, which have the potential to differentiate into both hepatocytic and cholangiocytic lineages. The clinical significance, as well as the biological behaviour, of CHC should be further characterized by more extensive studies and long-term follow-up reports.


This work was supported by grants from the Key Project of National Natural Science Foundation of China (Grant No. 30430670) and from the China Postdoctoral Science Foundation (Grant No. 20060400850).