Establishment of rat hepatocellular carcinoma cell lines with differing metastatic potential in nude mice



For better understanding of cancer metastasis, we have established an in vivo model for induction of highly metastatic hepatocellular carcinomas (HCC) in male F344 rats. From 1 tumor, 4 cell lines with differing metastatic potential (C1, C2, C6, C5F) were established by subcloning using the limited-dilution cloning technique. Two other lines, N1 and L2, arose from another primary HCC and a lung metastatic lesion, respectively. Although cell adhesion of each cell line in culture medium was different, tumors developing in the subcutis of nude mice after transplantation were all moderately differentiated HCC with a trabecular pattern. On subcutaneous injection into nude mice, all 6 cell lines proved to be tumorigenic in the injection site and C5F was highly metastatic to the lung. With injection into the tail vein, N1 and L2 formed frequent metastases in the lung as well as in lymph nodes. Using intraperitoneal injection, C1, C6, N1 and L2 showed marked disseminated growth in the abdominal cavity with bloody ascitis. Northern blot analysis revealed expression of known metastasis-related genes, KAI1 and heparanase, to be decreased in C5F, but no differences in expression of nm23-H1 were evident. A point mutation in the GSK-3β phosphorylation site of the β-catenin gene was found in L2. These transplantable HCC cell lines that have different metastatic ability should be useful for elucidation of mechanisms of metastasis. © 2001 Wiley-Liss, Inc.

Metastasis is a prominent phenotype of malignant tumors, along with autonomous proliferation and invasion, for which many steps involving accumulation of genetic changes and destruction of biological barriers are required. Altered expression of the putative metastasis suppressor gene KAI-11, 2 or nm23-H13, 4 is considered to play an important role during the acquisition of metastatic ability. Furthermore, change in cell-cell adhesion molecules, such as cadherins5, 6 and β-catenin,7 matrix metalloproteinases8 and heparanase,9, 10 and blood vessel adhesion factors11 may contribute to tumor metastasis.

Since metastasis is accomplished through dynamic changes that take place in the organism, it is very difficult to clarify the underlying mechanisms. We have developed an in vivo model in which chemically induced hepatocellular carcinomas (HCC) spontaneously metastasize to the lung at high incidence12, 13 and found that this reproducible system is able to provide a good tool to investigate steps in metastasis from the primary liver tumors to the lung. In the present experiment, for the further analysis of the biological alterations in this metastasis model, several HCC cell lines were established. Their growth and metastatic potential in nude mice as well as expression levels of known metastasis-related genes, KAI-1, nm23-H1 and heparanase, and mutations of β-catenin, H-ras and p53 genes were analyzed.


Tumor development

The original protocol for tumor induction was described in detail previously.12 In this experiment, a modified method was carried out as follows: 7-week-old male F344 rats (Charles River Japan, Atsugi, Japan) received an intraperitoneal (i.p.) injection of diethylnitrosamine (200 mg/kg b.w.; Tokyo Kasei Kogyo, Tokyo, Japan) and were then administered N-nitrosomorpholine (Tokyo Kasei Kogyo, Tokyo, Japan) mixed in the drinking water (120 ppm) for 16 weeks ad libitum. Rats were killed at experimental week 24 under anesthesia, and tissue samples were taken from liver tumors and metastatic lesions.

Primary HCC cell culture

For primary culture, 2 tumors from 1 rat and 1 lung metastatic nodule from another rat were removed (about 64 mm3 each), washed in Hanks' balanced salt solution and minced. They were then treated with 50 U/ml of dispase (Sanko Jyunnyaku, Tokyo, Japan) in Dulbecco's modified Eagle's medium (DMEM; GIBCO, Grand Island, NY) including 10% fetal bovine serum (FBS), 200 U/ml penicillin, 200 μg/ml streptomycin and 0.5 μg/ml amphotericin B (GIBCO) for 30–60 min at 37°C. Separated cells were incubated in DMEM additionally mixed with MITO (Collaborative Biomedical, Bedford, MA) for 30 min on dishes coated with collagen type I (Iwaki, Tokyo, Japan), then floating cell clusters including epithelial cells were transferred to new dishes. Epithelial cells were selected by differential attachment to plastic dishes during passage. From primary culture of 1 liver nodule, 4 subclones (C1, C2, C5F and C6) were isolated by limiting dilution using 98-well dishes. The other 2 lines, from another liver nodule (N1) and a lung metastasis (L2), were continuously passaged until mesenchymal cells were no longer apparent. After several passages, cells were cultured in DMEM without MITO in uncoated plastic dishes.

Small parts of the lesions, adjacent to the tissues removed for culture, were fixed in 10% formalin solution for the histological analysis. All 3 were diagnosed as moderately differentiated HCCs.

Inoculation to the nude mice

Exponentially growing cells were harvested with trypsin-EDTA, washed with DMEM and resuspended in FBS-free DMEM or Hank's balanced salt solution. For subcutaneous (s.c.) inoculation, cells (5 × 106/0.2 ml) were injected into the s.c. tissue of the left femurs of 8- to 10-week-old female athymic nude mice of the KSN strain (Nihon SLC, Hamamatsu, Japan). Tumor volume was estimated from 3 dimensions every week. All mice were killed 5 to 7 weeks after inoculation. For intravenous (i.v.) inoculation, cells (1 × 106/0.2 ml) were injected into the tail vein and the mice were killed 5 to 6 weeks thereafter. For i.p. inoculation, cells (1 × 106/0.2 ml) were injected into the lower right abdomen. Mice with marked ascitis were killed in a moribund condition, and other mice without any symptoms were killed 7 weeks after the injection.

At the sacrifice time point, visible tumors and the liver, lung, kidneys, spleen and lymph nodes of the pulmonary hilum and mesenterium were removed and fixed in 10% buffered formalin. At least 1 section of other tissues but 2 each for liver, kidney and the largest areas from each lobe of the lung were processed for HE staining. Quantitative analysis of lung metastases, to give numbers and areas (mm2) of lesions per cm2 of lung tissue, was carried out with the aid of an image processor (Image Processor for Analytical Pathology; Sumika Technoservise, Osaka, Japan)14 under the microscope.

Northern blotting for metastasis-related genes

Total RNAs from all cultured cell lines were extracted using TRIzol (GIBCO). Samples of the primary HCCs and lung metastasis nodules from the in vivo rat metastasis model and nontreated liver and lung tissues of F344 rats were also analyzed as controls. Probes were made by PCR based on the published cDNA sequences of rat KAI-1,2nm23-H1(NDPβ)3 and heparanase,9KAI-1 sense, 5′ TCC TCT TCC TCT TCA ATC TGC T 3′; and anti-sense, 5′ CCT CGT TTA CAG CAC CAA TAC A 3′; nm23-H1 sense, 5′ CTT GCA GCT GTA GGG AAC CA 3′; and anti-sense, 5′ CTG CTG CTC CCC TGC CTG TG 3′; and heparanase sense, 5′ GAA GGC TGG TGG AGA AGT GA 3′; and anti-sense, 5′ CCC AAT TTA TCC AGC CAC AT 3′. The sequences of purified PCR products were confirmed by auto sequencer (ABI PRISM; Applied Biosystems, Foster City, CA). Northern blotting was performed with a routine protocol.

SSCP analysis for β-catenin, H-ras and p53

DNA was also extracted from all culture cell lines using TRIzol, and PCR-SSCP analysis was performed to detect mutations of β-catenin with the primer sets described previously,15H-ras exon 1 with sense, 5′ GCG ATG ACA GAA TAC AAG CT 3′; and anti-sense, 5′ GAG CTC ACC TCT ATA GTG GG 3′; exon 2 with sense, 5′ TTG CAG GAC TCC TAC CGG AA 3′; and anti-sense, 5′ GAC TTG GTG TTG TTG ATG GC 3′; p53 exon 5 with sense, 5′ GAT TCT TTC TCC TCT CCT AC 3′; and anti-sense, 5′ AGT TCT AAC CCC ACA GCA GT 3′; exon 6 with sense, 5′ GCC TCT GAC TTA TTC TTG CT 3′; and anti-sense, 5′ AGT CTT CCA GCG TGA TGA TG 3′; and exons 7 with sense, 5′ TGC CTC CTC TTG TCC CGG GT 3′; and anti-sense, 5′ CCT CCA CCT TCT TTG TCC TG 3′. Five-microliter-volume aliquots including PCR buffer (Perkin Elmer, Foster City, CA), AmpliTaq (Perkin Elmer) and [α-32P]dCTP (Amersham, Arlington Heights, IL) were subjected to 40 cycles of amplification using a DNA Thermal Cycler (MJ Research, Waltham, MA). Each cycle consisted of 1 min denaturation at 94°C, 2 min reannealing at 57°C and 2 min of extension at 72°C for β-catenin and 45 sec denaturation at 94°C, 30 sec reannealing at 56/52/60/60/60°C (H-ras exons 1, 2, p53 exons 5, 6 and 7, respectively) and 1 min of extension at 72°C. Fifteen microliters of stop solution (95% formamide, 20 mM EDTA, 0.05% bromophenol blue, and 0.05% xylene cyanol) was then added to each reaction mixture. After 5 min of heat denaturing, 4 μl aliquots were applied to 0.5× MDE gels (AT Biochem, Malvern, PA). Electrophoresis was performed for 8 hr at 5 W and gels were dried and used to expose X-ray films. PCR products from the shifted bands were eluted, amplified with the same PCR primer sets and analyzed by uisng the Taq Dye Deoxy Terminator Cycle Sequence method.

Statistical analysis

To assess statistical significances, the 2-tail Student's t-test was used for the quantitative data and the χ2 test for the incidence data.


A total of 6 cell lines, 5 (C1, C2, C5F, C6, N1) from primary liver nodules and 1 (L2) from a lung metastasis, were established. In plastic culture dishes, their morphological characteristics differed (Fig. 1). C1, C2 and C6 cells were polygon-shaped and formed sheets by attachment with good adhesion to the surface. C5F formed aegagropila-like clusters, floating in the medium. N1 exhibited rounding and were attached loosely to the dish but as single cells. L2 exhibited a polygon shape with projections and attachment to both other cells and the dish. When confluent, cell clusters were also formed.

Figure 1.

Morphological appearance of each of the established rat HCC cell lines in culture medium.

All these cells could be maintained over 20 passages and proved tumorigenic in nude mice on s.c. injection (Table I). The subcutaneously formed tumors from all cell lines exhibited a similar histological character (Fig. 2), with trabecular patterns comparable with typical human HCCs. Central necrosis of subcutaneously growing tumors was frequently found when they became big. No significant difference in doubling time was evident among the cell lines.

Table I. Numbers and Areas of Lung Metastases From HCC Cell Lines in Nude Mice 5–7 Weeks After S.C. Injection
Cell lineNo. of miceS.c. tumorLung metastasis
Incidence (%)Incidence (%)Number (no./cm2)Area (mm2/cm2)
  • 1

    Significantly different at p < 0.05 vs. L2.

  • 2

    Significantly different at p < 0.05 vs. C2, C6, L2.

C133 (100)2 (67)17.5 ± 27.70.43 ± 0.64
C233 (100)1 (33)0.5 ± 0.90.03 ± 0.04
C688 (100)4 (50)1.4 ± 2.20.16 ± 0.37
C5F99 (100)8 (89)146.1 ± 56.521.44 ± 2.05
N11111 (100)7 (64)4.2 ± 5.90.15 ± 0.30
L255 (100)1 (20)1.0 ± 2.30.11 ± 0.24
Figure 2.

Histological appearance of tumors from the rat HCC cell lines subcutaneously transplanted into nude mice.

At sacrifice, no macroscopic metastatic nodules were noted in the lungs of nude mice receiving s.c. injection. Microscopically, however, there were many small lung metastatic lesions in the C5F-treated mice (Fig. 3). A few small foci were also evident in nude mice given the other cell lines. Data for the numbers and areas of metastatic HCC foci per cm2 lung section are summarized in Table I, significant differences being observed for number between the C5F and C2, C6 and L2 cases.

Figure 3.

Macroscopic (a–c) and microscopic (d–f) appearance of lung metastases in nude mice. Five to 7 weeks after s.c. injection of cell line C5F (a, d). Six to 7 weeks after i.v. injection of cell lines N1 (b, e) and L2 (c, f).

With the i.v. injection, some N1- or L2-treated mice died at week 5 with obvious lung and para-tracheal lymph node metastases (Fig. 3). Remaining animals of these groups were killed at weeks 6 or 7. In N1- and L2-treated mice, both numbers and areas of metastatic foci were significantly higher than in the C1, C6 and C5F cases (Table II). L2-treated mice had a tendency to develop bigger foci in the lung than those receiving N1.

Table II. Numbers and Areas of Lung Metastases From HCC Cell Lines in Nude Mice 6–7 Weeks After I.V. Injection
Cell lineNo. of miceLung metastasisLymph node metastasis
Incidence (%)Number (no./cm2)Area (mm2/cm2)Incidence (%)
  • 1

    Significantly different at p < 0.05 vs. C1, C6, C5F.

  • 2

    Significantly different at p < 0.05 vs. C5F.

  • 3

    Significantly different at p < 0.05 vs. C1, C2, C6, C5F, L2.

  • 4

    Significantly different at p < 0.05 vs. C1, C2, C6, C5F.

  • 5

    Significantly different at p < 0.05 vs. C6, C5F.

C130 (0)001 (33)
C232 (67)4.7 ± 5.52.4 ± 4.01 (33)
C682 (25)1.6 ± 2.90.8 ± 1.60 (0)
C5F80 (0)000 (0)
N11212 (100)148.8 ± 43.7322.1 ± 17.519 (75)5
L233 (100)215.3 ± 4.0125.7 ± 7.343 (100)5

With the i.p. injection, severe bloody ascites and peritoneal dissemination of HCCs were seen in C1-, C6-, N1- and L2-treated mice (Table III). The disseminated tumors were mainly observed in the great omentum and mesenterium (Fig. 4). Invasive growth into the liver capsule was noted in N1-treated mice. A few lung metastases were also evident with C6.

Table III. Incidence of Peritoneal Dissemination and Numbers and Areas of Lung Metastases from HCC Cell Lines in Nude Mice 3–7 Weeks After I.P. Injection
Cell lineNo. of micePeritoneal disseminationLung metastasis
Incidence (%)Incidence (%)Number (no./cm2)Area (mm2/cm2)
  • 1

    Significantly different at p < 0.05 vs. N1.

  • 2

    Significantly different at p < 0.05 vs. C1, C6, N1, L2.

C144 (100)0
C241 (25)10
C644 (100)2 (50)1.2 ± 1.70.02 ± 0.02
C5F4021 (25)8.5 ± 11.90.53 ± 0.75
N166 (100)0
L244 (100)0
Figure 4.

Macroscopic observation of abdominal organs 4 weeks after i.p. injection of cell line L2. Multiple white nodules are apparent on the peritoneum.

Compared with other cell lines, the expression levels of KAI1 and heparanase were obviously decreased in C5F, which showed high metastasis on s.c. injection, and were almost the same in the remaining cell lines, primary liver HCCs and normal liver. Expression levels of nm23-H1 were not altered in the 6 cell lines (Fig. 5).

Figure 5.

Northern blotting analysis of rat KAI1, heparanase and nm23-H1. Lane 1, normal lung; lane 2, normal liver; lanes 3–5, primary HCCs induced by diethylnitrosamine and N-nitrosomorpholine; lanes 6–8, lung metastases of HCCs; lanes 9–14, HCC cell lines; lane 9, C1; lane 10, C2; lane 11, C5F, lane 12, C6; lane 13, N1; lane 14, L2.

A point mutation of C (ACC; Thr) to T (ATC; Ile) in codon 41, in the GSK-3β phosphorylation consensus motif of rat β-catenin gene, was observed in L2 cells, established from a lung metastasis (Fig. 6). A point mutation of G (GAG; Glu) to A (AAG; Cys) in codon 76 of rat H-ras was detected in all cell lines and a point mutation T (ATC; Ile) to C (ACC; Thr) in codon 193 of the rat p53 was also noted in all but the L2 cell line.

Figure 6.

Results of SSCP (a) and sequence (b) analysis of β catenin. Shifted bands are evident with L2, and an ACC (Thr) to ATC (lle) point mutation is observed in codon 41.


In our study, 6 cell lines established from rat HCCs were clearly demonstrated to have widely differing metastatic ability in nude mice, despite a similar epithelial morphology. Historically, mouse B16 melanoma (origin in 1954 at the Jackson Laboratory), mouse Lewis lung carcinoma (established in 1951 by Dr. M.R. Lewis) and Dunning R-3327 rat prostatic adenocarcinoma (initially discovered in 1961 by Dr. W.F. Dunning16) cell lines have been frequently used for the metastasis-related experiments. It has been reported that orthotopical implantation in the corresponding organ of nude mice resulted in much higher metastatic rates.17, 18 From HCCs, Morris rat hepatoma,19 Yoshida rat ascitic hepatoma20 and Hep G2, Hep 3B human HCC21 cell lines have been widely used.

The examples described here have a similar genetic background, all from HCCs induced by the same treatment. Variation, however, was obvious in terms of the metastatic potential with s.c., i.v. or i.p. routes of administration. Metastasis from s.c. injection sites mimics all the steps of metastasis from the primary site to distant organs through vessels. In contrast, i.v. injection involves only late events, so that the number and size of metastatic nodules would be expected to be larger than with s.c. injection after the same duration. Interestingly, in our experiment, 2 cell lines (C1 and C5F) showed inconsistent behavior (Table IV). Especially in the case of C5F, s.c. injection caused abundant metastatic lesions, while no metastasis was observed on i.v. administration. The following explanations might be considered for this phenomenon. The primary s.c. tumor or tumor-bearing host animal may secrete factors that accelerate metastatic processes. Furthermore, variation in the flow pattern between s.c. and i.v., that is, limited but repeated release of cells from primary s.c. sites, may be more effective than a large number of cells given together in the i.v. case. Another possibility is that adhesion molecules on the surfaces of cells injected i.v. might be damaged by trypsinization.

Table IV. Metastatic and Peritoneal Disseminating Potential of the HCC Cell Lines
Cell lineOriginLung metastasisPeritoneal dissemination
C1Liver tumor 1++++
C2Liver tumor 1±+±
C6Liver tumor 1+++++
C5FLiver tumor 1+++
N1Liver tumor 2+++++++
L2Lung metastasis+++++++

However, the present data also revealed that the potential for peritoneal dissemination of tumor cells was not related to metastasis through blood vessels, as indicated in comparison with the s.c. and i.v. cases (Table IV). Attachment or invasion to endothelial cells and the peritoneum may be regulated independently in each of the cell lines.

Reduced expression of KAI-1 has been found to be associated with metastasis of cancers of the prostate,1 pancreas22 and the liver23 but not the esophagus or stomach.24 Decreased expression of another potent metastasis-suppressor gene, nm23-H1, was found in some25, 26 but not all27, 28 primary lesions of human metastatic HCCs. Recently, endoglycosidase heparanase, which cleaves a principal component of basement membranes, has been reported to be positively correlated with the metastatic potential in rat mammary adenocarcinoma cell lines,9 T-lymphomas, melanoma cell lines and human breast, colon and liver carcinoma specimens.10

In our present experiment, decreased expression of KAI-1 and heparanase were observed in C5F as compared with other cell lines. In samples taken from our in vivo rat model, they were relatively highly expressed in lung metastasis lesions but only rarely found in normal liver and primary HCCs. On the other hand, alteration of nm23-H1 expression was not observed in any cell line or tissue. Decreased expression of KAI-1 and heparanase in C5F might be related to the susceptibility of metastasis by s.c. and resistant of metastasis by i.v. and i.p., respectively, but none of these factors was clearly linked to the metastatic potential.

In our study, a hot spot-related mutation of the β-catenin gene was found in L2, established from a lung metastasis site. β-catenin was originally found as a molecule playing a fundamental role in the regulation of the E-cadherin-catenin cell adhesion complex.29, 30 It also functions in growth-signaling events through the Wnt-signaling pathway, inhibited by complex formation with APC and promoted by interaction with the DNA-binding proteins Tcf and Lef-1.31 Frequent mutation in the β-catenin GSK-3β phosphorylation consensus motif, which down-regulates binding to APC, has been reported in hepatocarcinogenesis in mice32, 33 and in the man34, 35 as well as rat and human colon carcinogenesis.15, 36, 37 Truncation of the β-catenin gene results in suppression of invasion and metastasis of human gastric cancer cells38 and human colorectal carcinomas,39 but there have been no reports of any relation between point mutations of β-catenin and metastasis. In our study, since growth potential did not differ among the cell lines, cell proliferation through the Wnt-signaling pathway does not appear to contribute in a major way to metastasis. On the other hand, cell adhesion ability associated with E-cadherin might be reduced in L2 and play a role in its relatively high metastatic ability. No mutations were seen in the hot spots of H-ras and p53, the similar mutation patterns presumably reflecting the genetic resemblance of these cell lines.

In conclusion, we established 6 HCC cell lines from lesions induced in vivo with a highly metastatic rat hepatocellular carcinoma model that formed histologically similar tumors when implanted s.c. The present genetic analysis could not explain the major differences in metastatic ability among the cell lines. But further investigations of mechanistic differences should cast light on metastatic processes.