Establishment and characterization of a novel primary hepatocellular carcinoma cell line with metastatic ability in vivo

Fresh tumor tissues from 30 HCC patients were included in the primary culture establishment study. After enzymatic digestion by type IV collagenase, disaggregated cells were collected from the tumors. Cell viabilities were assessed and only cases with cell viability >70% were subject to subsequent in vitro culture. For cases #1 to #20, 10 out of 20 cases (50%) generated cells with >70% viability. Primary cells usually require extracellular matrix components such as collagen and fibronectin or biodegradable polymers such as gelatin to promote cell attachment. Of these, gelatin and collagen were reported to favour sustained viability and functions of hepatocytes [16]. Therefore, freshly isolated cells were resuspended in AMEM supplemented with 10% FBS and then seeded onto 6-well culture plates coated with either gelatin or collagen. Cells from 5 out of 10 cases (50%) attached to the culture plates coated with 0.1% gelatin; while only 2 out of 10 cases (20%) attached to those coated with 0.1% collagen. Cells growing on collagen-coated plate were sparse and die within 1 month, while those growing on gelatin-coated plate attached better and could establish cell-to-cell contacts (Figure 1A). Therefore, 0.1% gelatin was chosen to coat plate for primary cell culture of subsequent HCC cases.

Fibroblast contamination and subsequent out-growth was frequently found in the process of cell line establishment. Therefore, we started to use defined media, hepatocyte culture medium (HCM), for cases #21 to #30. HCM provides a condition that permits preferential growth of hepatocytes, while that of contaminating fibroblasts is not favoured. For cases #21 to #30, 6 out of 10 cases (60%) generated cells with >70% viability, and the cells were resuspended in either AMEM supplemented with 10% FBS or HCM and then seeded onto 6-well culture plates coated with gelatin. Among the 6 cases, cells of #21 and #29 (2 out of 6 cases, 33%) could attach to culture plates coated with gelatin and grow in both 10% AMEM and HCM (Figure 1B). Fibroblast out-growth was observed in both cases cultured in AMEM with 10% FBS, while fibroblasts gradually decreased and disappeared in those cultured in HCM. Cells from #21 grew well and could propagate in subsequent passages, while those from #29 died gradually after one passage. Extensive characterization of cells from #21 was then performed.

We previously demonstrated that GEP was a hepatic oncofetal protein that defined CSC population in HCC [9]. CSCs are proposed to possess certain properties that allow them to survive better in primary culture [10]. Further analysis with clinico-pathological features revealed that increased GEP level was significantly associated with poor differentiation (p = 0.044) (Additional file 1: Table S1). As tumors with poor differentiation were more aggressive and associated with more stem-cell features [17], the result implied that GEP might contribute to the aggressiveness and stemness of HCC, and GEP-expressing cells might survive better in primary culture.

We quantified the GEP levels of the original tumors by flow cytometry, and studied their relationship with the cell viability and success rate of cell line establishment. GEP level was shown to be significantly correlated with the viability of cells isolated from the fresh tumor tissues (n = 28, Spearman’s ρ correlation coefficient = 0.729, p = 0.000) (Figure 2A). Cells isolated from 7 cases showed viability >70% and could successfully grow in culture, and their GEP levels were significantly higher than those that failed to grow or the viability was <70% (p = 0.001, Figure 2B). The results suggested that GEP might be an important factor for facilitating cell line establishment from fresh HCC tumor tissues.

During hepatocyte isolation, cells were detached from extracellular matrix. Such detachment induced anoikis and greatly decreases the cell viability and success rate of subsequent primary culture [5]. GEP was shown to protect cancer cells against anoikis [11],[12], and might therefore facilitate primary culture establishment by conferring anoikis resistance to the HCC cells. To support this hypothesis, Hep3B cells were sorted based on GEP expression. Post-sorting analysis was performed and purity was shown to be >80% (Additional file 1: Figure S1). Unsorted control cells, sorted GEP^high and GEP_low cells were then cultured in attachment free environment for 12 h. Viability of GEP_low cells was significantly lower than unsorted and sorted GEP^high cells (p = 0.025 and p = 0.011, respectively), suggesting that GEP was crucial for protecting HCC cells from anoikis-induced apoptosis (Figure 2C) and might facilitate primary culture establishment.

The cells derived from HCC case #21, designated HCC21, reached 80% confluency at 10 days after initial culture, and sub-passage was performed successfully without significant cell death. The cells began to grow quickly at the 6th passage and were passaged for more than 50 generations thereafter. Flow cytometric analysis showed that albumin + cells of the original tumor #21 were 83.2%. Upon culture for about 10 days in hepatocyte culture medium, albumin + cells increased to 88.9%, which was further enriched to more than 98% after 6 passages (Figure 3A), confirming that HCC21 cells were hepatocytes. HCC21 cells grew as adherent monolayer with epithelial morphology (Figure 3B) and the cells maintained consistent morphology along passages. Growth curve of HCC21 cells is shown in Figure 3C. The population doubling time of HCC21 was approximately 95 h. Wound healing assay was performed to assess the migration ability of HCC21 cells. The cells started to migrate a day after the wound was made, and the wound healing was completed after 3 days (Figure 3D). The migration ability of HCC21 cells echoed the histology observation that venous infiltration (micrometastasis) was observed in the surgical specimen.

The DNA samples of HCC21 early and late passages (11 and 50, respectively), original tumor and adjacent non-tumor liver tissue were subjected to DNA fingerprinting analysis. A total of 15 STR loci (CSF1P0, D2S1338, D3S1358, D5S818, D7S820, D8S1179, D13S317, D16S539, D18S51, D19S433, D21S11, FGA, TH01, TPOX, vWA) were co-amplified in each sample. AMLEO locus at the sex chromosomes was also examined. The data were analyzed and allele(s) of each locus were determined (Table 1). The STR profiles of HCC21 cells and the original tumor were identical, suggesting that they came from the same person. Besides, loss of heterozygosity (LOH) was observed in 2 loci, D16S539 and FGA in original tumor, and the aberrations were retained in HCC21 cells (Figure 4).

Changes in genomic copy number in HCC21 cells and original tumor were characterized by aCGH analysis. aCGH arrays were performed using DNA from original tumor and that from the HCC21 cells 6 months after its establishment. Chromosomal aberrations including chromosomal loss at 1p35-p36, 1q44, 2q11.2-q24.3, 2q37, 4q12-q13.3, 4q21.21-q35.2, 8p12-p23, 15q11.2-q14, 15q24-q26, 16p12.1-p13.3, 16q, 17p, 22q and gain at 1q21-q43 were observed in both the original tumor specimen and HCC21 cells (Figure 5). Noted the STR profiles showed LOH at FGA (at chromosome 4q28) and D16S539 (at chromosome 16q24.1), and that corroborated the aCGH data with chromosomal loss at 4q21.21-q35.2 and 16q in both the original clinical tumor specimen and the derived HCC21 cells.

p53 has been reported to be frequently mutated and overexpressed in HCCs [18]. Result from western blot showed that p53 protein was present in HCC21 cells. Human liver cancer cell lines HepG2 and PLC/PRF/5 were included as wild-type and mutant controls, respectively (Figure 6A). Subsequent analysis by DNA sequencing indicated a point mutation (Gly → Arg) (GGA → AGA) at codon 266 of exon 8 in both the HCC21 cells and tumor specimen (Figure 6B). PLC/PRF/5 was included as p53 mutant control (codon 249 of exon 7) and no mutation was found in exon 7 of HCC21 cells (data not shown).

HCC21 cells were injected subcutaneously into 4 nude and 3 NOD/SCID mice. About 2 months after injection, visible tumors developed in all mice at the site of inoculation (Figure 7A), indicating that HCC21 cells were tumorigenic. We assessed the expression of HCC markers including AFP and HBV antigen [19] in the HCC21 cells and xenograft tumors. Western blot analysis showed that HCC21 cells and xenograft tumors in both nude and NOD/SCID mice retained the expression of HBV core antigen and AFP, which was consistent with original resected tumor (Figure 7B). Most importantly, secondary tumors were found in the peritoneal cavities of 2 NOD/SCID mice (Figure 8A). Histological sections of the xenografts showed neoplastic hepatocytes proliferation in broad trabeculae and acinar pattern, morphologically consistent with HCC. In addition, the metastatic (secondary) lesions were morphologically similar to the primary tumors (Figure 8B, upper panel). Besides, we have examined the expression of AFP in the primary and secondary xenograft tumors, in comparison to the patient’s tumor and adjacent non-tumor liver tissue specimen by immunohistochemical staining. Results showed that AFP protein was expressed at high levels in the tumor specimen, and also in the primary and secondary xenograft tumors. AFP protein could also be detected in adjacent non-tumor liver tissue specimen, but the signal was lower than the tumor counterpart (Figure 8B, middle panels). This further demonstrated that the tumor characteristics had been preserved from the original tumor specimen to the primary and metastasized tumor xenografts.

The aCGH analysis revealed loss of 16q in HCC21 cells, which region contained E-cadherin gene (at chromosome 16q22.1). E-cadherin is an adhesion molecule crucial for inhibiting metastasis by connecting homophilic cells [20]-[22]. Loss of E-cadherin was reported to associate with high incidence of lymph node metastasis in various cancers, including HCC [23]-[25]. Therefore, we examined the expression of E-cadherin in HCC21 cells by western blot. HepG2, HCC cell line without metastatic potential, was included as positive control for the presence of E-cadherin. Result showed that E-cadherin protein was absent in HCC21 cells (Figure 8C), thereby providing further evidence for the metastatic potential of the cells.

The study protocol was approved by the Institutional Review Board of the University of Hong Kong / Hospital Authority Hong Kong West Cluster (HKU/HA HKW IRB). Between July 2010 and March 2011, 30 patients having curative partial hepatectomy or liver transplantation for HCC at Queen Mary Hospital, Hong Kong, were recruited after written informed consent was obtained. Tumors and non-tumorous tissues were collected from the resected specimens. A portion of tumor tissues was freshly processed for cell culture and flow analysis, while DNA and proteins were extracted from the snap frozen tissues.

Original tumor tissue from which HCC21 cells was derived was obtained from a 42-year-old Hong Kong Chinese female patient who underwent curative partial hepatectomy. Tumor size was 8.5 cm in diameter. The patient was seropositive for hepatitis B virus (HBV), and serum α-fetoprotein (AFP) level was 30690 ng/mL (reference range <10 ng/mL). A diagnosis of HCC arising in a cirrhotic liver was confirmed by histological examination. Venous infiltration was observed, indicating the aggressiveness of the tumor. The tumor was classified as stage II according to the pathological tumor-node-metastasis (pTNM) staging system 2009 version, and graded as moderately differentiated.

Fresh tumor tissues were rinsed with AMEM medium (Invitrogen, Carlsbad, CA) supplemented with 50 units/ml Penicillin G (Invitrogen) and 50 μg/ml streptomycin (Invitrogen), and then minced into 1-mm³ pieces. After digestion with type IV collagenase (Sigma, St. Louis, MO) for 5 min at 37°C for 3 times, disaggregated cell suspension was obtained by filtering through a 40 μm cell strainer (BD Biosciences, San Jose, CA). After lysis of red blood cells by ACK lysis buffer (Invitrogen), cells were washed with AMEM medium. The cells were then resuspended in either AMEM supplemented with 10% FBS (Invitrogen) or hepatocyte culture medium (Lonza, Basel, Switzerland), and then seeded onto 6-well culture plates coated with either 0.1% gelatin (Sigma) or 0.1% type I collagen (Sigma). Cells were incubated at 37°C in a humidified atmosphere at 95% air and 5% CO₂ and left untouched for 2 days. The culture medium was then changed twice a week and cells were sub-passaged when they reached 70-80% confluency.

For HCC21 cells, a split ratio of 1:1 was applied in the early passages (passage 1 to 5), thereafter increased to 1:3. Cells were collected at different passages and put in freezing medium containing 50% hepatocyte culture medium, 40% FBS and 10% DMSO, and stored in liquid nitrogen. The cells were tested for mycoplasma contamination and the result was negative.

Hep3B cells were sorted for GEP^high and GEP_low subpopulations by magnetic activated cell sorting (Miltenyi Biotec) as previously described [9]. After cell isolation, one million of each sorted population was collected to assess cell viability and purity by trypan blue staining and flow cytometry, respectively. Post-sorting analysis typically indicated purities of at least >80% with minimal cell death (<10%). Unsorted control cells, sorted GEP^high and GEP_low cells were then cultured in 10% FBS-supplemented AMEM in ultra-low attachment plate (Corning Inc, Corning, NY) for 12 h. Cells were harvested and stained for annexin V-propidium iodide (BD Biosciences) for viability, in which cells negative for both annexin V and propidium iodide were defined as viable cells. Viable cells were quantified by flow cytometric analysis.

Cells were permeabilized with ice-cold 0.1% saponin and then incubated with FITC-conjugated mouse anti-human GEP antibody (described previously [53]) (Versitech Ltd, Hong Kong), mouse anti-human albumin (R&D systems, Minneapolis, MN) or equal amount of mouse IgG isotype (Sigma). Cells were washed with 0.1% saponin and then subject to flow cytometric analysis. Results were expressed as percentage of positive cells or mean fluorescence intensity (MFI), after subtracting the non-specific background signal (isotype control).

Cells were routinely monitored and photographed with a phase-contrast microscope. Cells of passages 6 and 16 were studied to measured the population doubling time, which was assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay for 5 consecutive days.

Cells were seeded onto a 6-well culture plate and incubated for 24 hr. A wound was made by scraping a 20 μl pipette tip across the cell monolayer. Cells were rinsed with PBS and cultured in HCM for 3 days. Cell movement toward the wound was observed under a phase-contrast microscope every 24 h after the wound was made.

Genomic DNA was extracted from HCC21 cells and the original tumor. The Agilent human 44 K CGH microarray (Agilent Technologies, Santa Clara, CA) with an average probe spatial resolution of ~ 35 kb was utilized. Genomic DNA at 1 μg was labelled by the Agilent Genomic DNA Labeling Kit PLUS (Agilent Technologies). The hybridized array was scanned by an Agilent Microarray Scanner and data were analysed by Agilent Feature Extraction 9.1 followed by computation and normalization using CGH Analytics v3.4. Putative chromosome copy number loss was defined by intervals of two or more adjacent probes with log₂ ratios suggestive of a deletion when compared with the log₂ ratios of adjoining probes. The Quality-Weighted Interval Score algorithm (ADM2) with a value of 6.0 was used to compute and assist the identification of aberrations for a given sample.

Total protein was extracted with cell lysis buffer (Cell Signaling Technology, Boston, MA) in the presence of complete protease inhibitor cocktail (Roche, Mannheim, Germany) and separated in 8-10% SDS PAGE gel. Proteins were then electro-transferred onto polyvinylidene difluoride membranes, subsequently incubated with primary anti-human antibodies, detection by horseradish peroxidase-labeled secondary antibodies, and visualized with Enhanced Chemiluminescence Western Blotting Detection Kit (Amersham Biosciencse, Piscataway, NJ).

Direct DNA sequencing was performed for exons 4-9 of p53, in which >80% of all mutations were observed [54],[55]. Primer sets and reaction conditions were adopted from Lehman et al[56]. DNA was amplified by polymerase chain reaction and direct DNA sequencing was performed with the BigDye Sequencing kit (Invitrogen). Electrophoresis and sequence analysis were performed using the ABI PRISM 3100 (Invitrogen).

The study protocol was approved by and performed in accordance with the Committee of the Use of Live Animals in Teaching and Research at the University of Hong Kong. Cells from passage 8 were harvested, washed, and resuspended in hepatocyte culture medium (HCM). 1x10⁶ cells were injected subcutaneously into the right flank of each athymic nude or NOD/SCID mouse (4 weeks old). The mice were examined every week for the development of tumors and tumor-bearing mice were sacrificed when tumor burden exceed 10% of the normal body weight or the body weight loss exceed 20%.

Immunohistochemistry was performed with the Dako Envision Plus System (Dako, Carpinteria, CA) following the manufacturer’s instruction with modifications. Briefly, antigen retrieval was performed by microwave with sections immersed in citrate buffer. Followed by endogenous peroxidase blocking, tissues were stained with rabbit anti-human AFP antibody (Dako) or rabbit Ig (R&D Systems, Minneapolis, MN). The signal was detected by horseradish peroxidase-conjugated secondary antibody and color was developed with diaminobenzidine as the chromogen. The tissue sections were then counterstained with hematoxylin.

All data were expressed as mean values + standard deviation (SD) from at least three independent experiments. Differences between groups were assessed by the Student’s t test. A probability (p) < 0.05 was considered significantly different. All analyzes were performed using the statistical software GraphPad Prism for Windows, Version 3.00 (GraphPad Software, CA).