Hepatocellular carcinoma development induced by conditional β‐catenin activation in Lkb1 ^+/− mice

Abstract

The development of hepatocellular carcinomas (HCC) appears to be a multistep process that takes several decades in humans. However, the identities of specific gene alterations and their contribution to HCC pathogenesis remain poorly understood. We previously reported that Lkb1 ^+/− mice spontaneously develop multiple hepatic nodular foci (NdFc) followed by HCC, and that the conditional activation of β‐catenin in Catnb ^lox(ex3) mouse livers alone does not cause tumor formation. We show here that the conditional activation of β‐catenin accelerates HCC development in Catnb ^+/lox(ex3) Lkb1 ^+/− compound mutant mice, affecting displastic hepatocytes in NdFc that suffered LOH at the Lkb1 locus. We further show that β‐catnin activation provides HCC with a growth advantage as well as transplantability. These results suggest that the loss of Lkb1 contributes to the formation of dysplastic NdFc, and that Wnt signaling activation is involved in ensuing progression toward HCC. A combination of these sequential changes can be a practical model for a subset of human HCC. (Cancer Sci 2009)

Peutz–Jeghers syndrome (PJS) is an autosomal dominant disease characterized by gastrointestinal hamartoma and mucocutaneous pigmentation as well as the increased risk of cancer.^{( 1} , ^{2 )} The positional cloning studies on chromosome 19p13.3 identified mutations in LKB1 responsible for PJS.^{( 3} , ^{4 )} LKB1 is a serine/threonine kinase that phosphorylates and activates 14 kinases, including AMP‐activated protein kinase (AMPK) and microtubule‐associated protein/microtubule affinity‐regulating kinases (MARK).^{( 5 )} Recent studies have revealed that LKB1 mediates energy‐dependent suppression of the mammalian target of rapamycin (mTOR) pathway via AMPK,^{( 6 )} and that LKB1 suppresses tubulin polymerization by activating MARK–microtubule‐associated protein signaling.^{( 7 )}

We previously reported that Lkb1 ^+/− mice develop gastrointestinal hamartomas that have similar histopathology to those in human PJS.^{( 8 )} In addition, they developed multiple small nodular foci (NdFc) in the liver after 30 weeks of age, and HCC after 40 weeks of age.^{( 9 )} NdFc consist of dysplastic and proliferative hepatocytes, suggesting that HCC in Lkb1 ^+/− mice originate from these lesions.^{( 9 )} LKB1 mutations have been identified not only in PJS patients but also in sporadic cancers.^{( 10 )} Although the liver malignancy in PJS patients has been rarely reported, LKB1 is implicated in the pathogenesis of human HCC. A HCC cell line (BEL9204) and an immortalized hepatocyte line (TPH1) were found to bear homozygous deletions of the LKB1 gene,^{( 11 )} and 1 of the 80 primary HCC examined had a point mutation in the LKB1 kinase domain.^{( 12 )} These results suggest that LKB1 deficiency may underlie a subset of human HCC.

Wnt signaling activation is a frequent change in HCC as well as in other cancers.^{( 13 )} Upon activation, the adenomatous polyposis coli (APC)–axin–glycogen synthase kinase 3β (GSK3β) complex is unable to phosphorylate β‐catenin, which results in the stabilization of β‐catenin followed by the activation of T‐cell factor (TCF)/lymphoid enhancer factor (LEF)‐dependent transcription. β‐Catenin mutations in GSK3β phosphorylation sites have been identified in human, mouse, and rat HCC.^{( 14} , ¹⁵ , ^{16 )} We previously constructed Catnb ^lox(ex3) mice, in which exon 3 of the gene encoding β‐catenin was sandwiched by two loxP sites.^{( 17 )} Liver‐specific recombination by Cre recombinase resulted in hepatomegaly with mitochondrial swelling in hepatocytes.^{( 18 )} However, no neoplastic changes of hepatocytes were observed as a result of the activation of β‐catenin for up to 6 months following Cre induction by the recombinant adenovirus (AdCMV‐cre). On the other hand, another Wnt signaling activation model was reported in which the Apc gene was conditionally disrupted by Cre recombinase.^{( 19 )} Although these mice developed HCC and preneoplastic foci, it took 8–9 months after Cre induction, and latency was too long despite efficient recombination in approximately half of all hepatocytes. These results suggest that Wnt signaling activation alone is insufficient for hepatocarcinogenesis, in contrast to the extensive polyposis in the intestines of Catnb ^lox(ex3) Krt19 ^cre and Apc ^Δ716 mice.^{( 17} , ^{20 )}

It appears that single gene mutations are not sufficient to cause HCC, as various gene mutations and chromosomal aberrations are accumulated during multistep hepatocarcinogenesis in humans.^{( 21 )} To investigate the possible cooperation between the loss of LKB1 and the activation of β‐catenin in HCC development, we introduced the Catnb ^lox(ex3) allele to Lkb1 ^+/− mice and conditionally activated Wnt signaling in their livers by AdCMV‐cre infection.

Materials and Methods

Mice.  Construction of Lkb1 knockout and Catnb ^lox(ex3) mice has been described previously.^{( 8} , ^{17 )} We used only males due to the low incidence of NdFc and HCC in female Lkb1 ^+/− mice, as previously reported.^{( 9 )} For the adenoviral infection study, we generated Catnb ^lox(ex3) , Lkb1 ^+/−, and the compound mutant mice by intercrossing Catnb ^lox(ex3) and Lkb1 ^+/− mice. We further used age‐matched Lkb1 ^+/− mice to obtain additional NdFc and HCC samples for the following analyses. To inoculate and maintain transplanted HCC, we purchased C.B‐17/lcr‐scid/scid (Fox Chase SCID) and BALB/cA‐nu/nu (Nude) mice from CLEA Japan (Tokyo, Japan). All animal experiments were approved by the Animal Care and Use Committee of Kyoto University.

LOH analysis.  DNA extraction from paraffin‐embedded sections and genomic PCR for the targeted and wild‐type Lkb1 alleles were carried out as described previously.^{( 22 )}

Antibodies.  Antibodies for GS (FL‐373), p27 (C‐19), cyclin A (H‐432), cyclin D1 (M‐20), cyclin E (M‐20; Santa Cruz Biotechnology, Santa Cruz, CA, USA), human LKB1 (sheep polyclonal), mouse LKB1 (rabbit polyclonal), c‐Myc (rabbit polyclonal; Upstate, Charlottesville, VA, USA), β‐catenin (rabbit polyclonal; Sigma, St Louis, MO, USA), Ki‐67 (TEC3; Dako, Glostrup, Denmark), and GAPDH (6C5; Ambion, Austin, TX, USA) were purchased from commercial sources.

Western blotting analysis.  Sample preparations from tissues, electrophoresis, and transfer to nitrocellulose membrane were carried out as described previously.^{( 8} , ^{9 )} Membranes were incubated with primary antibodies diluted in Blocking One (Nacalai Tesque, Kyoto, Japan) at 4°C overnight and secondary antibodies at room temperature for 30 min. Immobilon Western HRP Chemiluminescent Substrate (Millipore, Bedford, MA, USA) was used to detect the specific signals. Band intensity was visualized and quantified using Image Gauge application software (Fujifilm, Tokyo, Japan).

Histopathological analysis.  Preparation of histological specimens and immunostaining were carried out as described previously.^{( 8} , ^{9 )} Antigen retrieval by microwave was carried out in 10 mM citrate buffer (pH 6.0) before incubation with the primary antibody for Ki‐67. Brightfield images were captured and analyzed with Nikon ECLIPSE E800 (Tokyo, Japan) and Adobe Photoshop software (San Jose, CA, USA).

Microarray analysis.  Total RNA was extracted from tumors by RNeasy Plus (Qiagen, Valencia, CA, USA). Cy3‐ and Cy5‐labeled amplified antisense RNA (aRNA) was synthesized by the Amino Allyl Message Amp II aRNA Amplification Kit (Ambion). A mixture of two‐colored probes was competitively hybridized to the arrays spotted with the mouse oligo library (Sigma). After washing, arrays were dried and scanned by the microarray scanner system G2565BA (Agilent Technologies, Palo Alto, CA, USA). The data were analyzed using Feature Extraction software (Agilent Technologies).

RT‐PCR.  Total RNA was prepared from tissues using ISOGEN solution (Nippon Gene, Tokyo, Japan), and cDNA was synthesized with SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA, USA). Primer sequences for specific gene amplification are shown in Table S1.

Human HCC samples.  Surgical specimens of the primary HCC were collected, with informed consent, upon surgery at Kyoto University Hospital with the protocol approved by the Ethics Committee of Kyoto University.

Results

Loss of Lkb1 heterozygosity causes dysplastic NdFc in the liver of Lkb1 ^+/− mice.  We previously reported that all HCC in the Lkb1 ^+/− mice tested had lost expression of LKB1 protein due to the loss of Lkb1 heterozygosity.^{( 9 )} To investigate whether dysplastic cells in the NdFc of Lkb1 ^+/− mice also show LOH at the Lkb1 locus, we carried out PCR using genomic DNA samples from the NdFc and adjacent normal tissues, separated by laser microdissection. In all of the 14 pairs of samples, the band intensities for the wild‐type allele were significantly weaker in the NdFc than in the adjacent normal tissues (Fig. 1 and data not shown), suggesting that loss of LKB1 expression by LOH at the Lkb1 locus is likely an initiating event of hepatocarcinogenesis in Lkb1 ^+/− mice.

Figure 1.

 Loss of intact Lkb1 allele in preneoplastic nodular foci (NdFc) of Lkb1 ^+/− mice. A genomic PCR analysis for Lkb1 targeted alleles (KO) and wild‐type alleles (Wild) was carried out using template DNA from four pairs of NdFc (Fc) and adjacent normal tissues (N). As a control, a pair of fibroblast (MEF) preparations from Lkb1 ^+/− and Lkb1 ^−/− embryos was also examined.

Activation of β‐catenin accelerates hepatocarcinogenesis in Catnb ^+/lox(ex3) Lkb1 ^+/− compound mutant mice.  To investigate possible cooperation between the loss of LKB1 and the activation of β‐catenin in HCC development, we introduced a floxed stabilizing β‐catenin gene (i.e. Catnb ^lox(ex3) allele) into the Lkb1 ^+/− mice and generated Catnb ^+/lox(ex3) Lkb1 ^+/− mice (hereafter compound mutant mice). Expression of β‐catenin^Δex3, a stabilized form lacking the GSK3β phosphorylation sites, can be induced in hepatocytes of these mice by Cre‐mediated excision of exon 3 from the Catnb ^lox(ex3) allele through intravenous injection of the Cre‐expressing adenovirus AdCMV‐cre.^{( 17 )} Because high multiplicity of infection of the hepatocytes with AdCMV‐cre causes fatal Wnt signaling,^{( 18} , ^{19 )} we adjusted the adenovirus titer to 2 × 10⁷ pfu per mouse, a dose that should activate Wnt signaling only in ∼10% of the hepatocytes (Fig. S1). Catnb ^+/lox(ex3) (hereafter Catnb ^lox(ex3) mice), Lkb1 ^+/−, and their compound mutant mice were infected with AdCMV‐cre at 11–14 months of age. They were euthanized 2 months post‐infection and their livers were examined (Fig. 2a). Two of seven Lkb1 ^+/− mice developed single detectable tumors on the liver surface, whereas only one of seven Catnb ^lox(ex3) mice had a tumor that was diagnosed as NdFc histologically (Fig. 2b,c). On the other hand, the compound mutant mice developed multiple tumors, ranging from 0 to 14 per animal, that were visible on the liver surface. These results suggest that the activation of β‐catenin accelerates hepatocarcinogenesis in Lkb1 ^+/− mice. In contrast, no macroscopic liver tumors were found when younger compound mutant mice at 2 months of age were Wnt‐activated in the same manner (data not shown). Considering the long latency of hepatocarcinogenesis in Lkb1 ^+/− mice (Fig. 2a), it was likely that liver tumors in the compound mutant mice derived from microscopic NdFc rather than from normal hepatocytes. It was also possible, however, that activation of β‐catenin increased the number of NdFc in the aged compound mutant mice, which could have accelerated the liver tumor development by expanding the NdFc population. To test this possibility, we estimated the number of microscopic foci per unit volume of the liver using a mathematical formula.^{( 23 )} However, the number of nodules estimated in the compound mutant mice was even lower than that in the Lkb1 ^+/− mice (Fig. 2d), excluding the possibility that activation of β‐catenin was involved in the early events of hepatocarcinogenesis in the compound mutant mice.

Figure 2.

 Acceleration of hepatic tumorigenesis by activation of β‐catenin in Catnb ^+/lox(ex3) Lkb1 ^+/− compound mutant mice. (a) Experimental schedule for this study. All animals in the crossing experiment were infected with AdCMV‐cre at the ages shown in months (mo). (b) Tumor multiplicities 2 months after AdCMV‐cre infection. Tumors larger than 3 mm in diameter visible on the liver surface were counted under a dissection microscope. Statistical analysis was carried out with Mann–Whitney’s U‐test. *P < 0.05. One tumor that was found in a Catnb ^lox(ex3) mouse (open circle) was a NdFc histologically. (c) Gross appearance of representative liver tumors (arrowheads). †Connective tissue associating with vessels. Scale bars = 1 cm. (d) Estimated numbers of microscopic NdFc in the liver. The number of microscopic nodules larger than 200 μm in diameter per cm³ liver was calculated as described.^{( 23 )} Data are expressed as the mean with SD. The difference of average NdFc numbers between Lkb1 ^+/− and the compound mutant mice was not statistically significant by Student’s t‐test. (e) SCID mice inoculated with primary tumors from a Lkb1 ^+/− mouse (left) and a compound mutant mouse (right). Photographs were taken 2 months after transplantation. #Tumor inoculation site.

HCC in the compound mutant mice have higher transplantability than those in Lkb1 ^+/− mice.  To test the effect of the activation of β‐catenin on transplantability of the HCC, we subcutaneously inoculated the primary HCC clumps from the Lkb1 ^+/− and compound mutant mice into immunocompromised SCID mice. Although no growths were obtained in 2 months from any of the four HCC derived from Lkb1 ^+/− mice, two of three HCC from the compound mutant mice formed tumors of >1 cm in diameter (Fig. 2e). After long‐term incubation (>6 months), we could establish two transplantable HCC lines from the Lkb1 ^+/− mice as well. These results indicate that activated β‐catenin increases the transplantability of HCC in immunocompromised mice.

Concomitant loss of LKB1 and Wnt signal activation are reflected in the transcriptomes of HCC of compound mutant mice.  We determined the expression levels of LKB1 and β‐catenin protein in Lkb1 ^+/− and the compound mutant mouse livers (Fig. 3). HCC of Lkb1 ^+/− mice expressed a similar level of β‐catenin to non‐tumor tissues, suggesting that Wnt signaling was not activated in these tumors. Truncated β‐catenin was detected in both HCC and normal adjacent tissues of the compound mutant mice (Fig. 3). We found that all HCC in the compound mutant mice lacked expression of LKB1 as in Lkb1 ^+/− mice, and PCR analysis using the laser microdissection samples from those HCC showed that they had undergone LOH at Lkb1 loci (data not shown). Interestingly, HCC in Lkb1 ^+/− mice showed Akt activation whereas the compound mutant mice did not. These results collectively suggest that the activation of β‐catenin promotes tumor progression in the dysplastic hepatocytes that had lost LKB1, through a mechanism distinct from Lkb1 ^+/− mice.

Figure 3.

 Concomitant changes of LKB1 and β‐catenin expression in liver lesions. Western blotting analysis for β‐catenin (WT, wild‐type; Δex3, stabilized mutant), LKB1, phospho‐Akt (Ser473), and total Akt with GAPDH control is shown. A wild‐type liver (W), and pairs of tumor (T) and adjacent normal tissue (N) from two Lkb1 ^+/− and two compound mutant mice were analyzed.

To obtain insights into the effects of β‐catenin activation on the transcriptome of HCC, we next compared the gene expression profiles of transplanted HCC (C1 and C2 from the compound mutant mice and L1 and L2 derived from the Lkb1 ^+/− mice) for 22 056 mRNA species, using two sets of microarray data (C1 vs L1 and C2 vs L2). We listed genes whose expression levels changed more than four‐fold in both assays, and found that 36 genes were upregulated whereas 24 genes downregulated in C1 and C2 compared with L1 and L2 (Tables S2 and S3, respectively). Table 1 summarizes the representative genes that were upregulated and downregulated in transplants with the activation of β‐catenin. We further picked up several upregulated genes from the list and confirmed the microarray data by RT‐PCR using cDNA from transplanted tumor lines (Fig. S2). Although expression of glutamine synthetase (GS) was not upregulatad (Fig. S3 and data not shown), the list includes several targets of the Wnt–β‐catenin signaling pathway such as Oat, Lect2, and Lgr5.^{( 26} , ²⁷ , ^{28 )} Therefore, we concluded that the canonical Wnt signaling pathway was activated in HCC of the compound mutant mice but not in those of Lkb1 ^+/− mice.

Table 1.

 Upregulated and downregulated genes in hepatocellular carcinomas (HCC) with the activation of β‐catenin

Category and symbol	Fold change†		Reference‡
	C1/L1	C2/L2
Transporter
Slco1b2	66.15	67.09
Slc10a1	18.72	20.20
Slc21a5	8.38	6.01
Slc1a2	7.71	8.59	(24)
Rhbg	6.68	6.10	(24, 25)
Slc16a7	4.24	6.99	(24)
Abcc1a	0.23	0.24
P450
Cyp2e1	23.43	19.32	(25)
Cyp2c39	11.57	17.36
Cyp2c29	8.96	10.02	(25)
Cyp1a2	4.59	17.62	(25)
Cyp27a1	4.24	10.14
Cyp2f2	0.14	0.05	(24)
Amino acid metabolism
Oat	8.97	7.98	(25, 26)
Asns	7.07	7.52
Hal	0.21	0.04	(24)
Sds	0.22	0.09	(24)
Secretory protein
Tgn	20.12	19.37
Spp2	14.46	7.25
Igfbp2	10.57	4.46
Lect2	8.86	17.02	(27)
Other
Lgr5	10.80	8.48	(24, 28)
2310007H09Rik	9.47	6.75	(24)
Avpr1a	7.69	11.85	(24)
Raet1c	0.05	0.08

†Fold changes in the transplanted HCC of the compound mutant mice (C1 or C2) compared with those of Lkb1 ^+/− mice (L1 or L2) are shown. ‡References are shown, in which the correlation with the Wnt signaling is reported.

HCC in Lkb1 ^+/− and compound mutant mice have similar histological phenotypes.  We next studied the histopathology of the liver lesions in the compound and simple mutant mice. Wild‐type livers showed well‐organized hepatic cords (Fig. 4a), whereas NdFc in Lkb1 ^+/− mice and Catnb ^lox(ex3) mice were composed of large dysplastic hepatocytes lacking the cord structure (Fig. 4b,c). Most HCC found in either Lkb1 ^+/− or the compound mutant mice were of the trabecular type, consisting of well‐differentiated carcinoma cells that formed disorganized cord structures (Fig. 4d,e). Three of 17 HCC in the compound mutant mice also contained a gland‐like structure (data not shown). Together with our previous report,^{( 9 )} these results show that the histopathological features of HCC in the compound mice are similar to those observed frequently in Lkb1 ^+/− mice.

Figure 4.

 Minimal effects of activated β‐catenin on histopathology of hepatocellular carcinomas (HCC). HE stains of (a) wild‐type liver, (b) a nodular focus in a Catnb ^lox(ex3) mouse, (c) a nodular focus in a Lkb1 ^+/− mouse, (d) a HCC in a Lkb1 ^+/− mouse, and (e) a HCC in a compound mutant mouse. Insets are at a higher magnification. Arrowheads show the boundary between normal and tumor tissues. Scale bars = 500 μm.

Wnt signal activation increases growth rate in HCC.  To assess the effects of Wnt signal activation on the liver lesions, we immunostained proliferative cells for Ki‐67 protein. It is expressed in all cycling cells,^{( 29 )} and was found in the NdFc and HCC of Lkb1 ^+/− mice, HCC of the compound mutant mice, and NdFc of Catnb ^lox(ex3) mice (Fig. 5a–d). The fraction of Ki‐67‐positive cells was ∼20% in HCC of the compound mutant mice, which was significantly higher than in Lkb1 ^+/− mice as well as in the NdFc of Lkb1 ^+/− or Catnb ^lox(ex3) mice (Fig. 5e). A similar increase in growth rate was observed in the HCC transplanted from the compound mutant mice (C1 and C2) compared with those in the Lkb1 ^+/− mice (L1 and L2) (Fig. 5f). These results suggest that the increase in the proliferating cell fraction contributes to the rapid HCC tumorigenesis after infection of the compound mutant mice with AdCMV‐cre.

Figure 5.

 Increase in the proliferating cell fraction in hepatocellular carcinomas (HCC) with activated β‐catenin. Immunostaining for Ki‐67 antigen in a (a) nodular focus and (b) HCC of Lkb1 ^+/− mice, (c) a nodular focus of a Catnb ^lox(ex3) mouse, and (d) a HCC of a compound mutant mouse. Scale bars = 200 μm. (e) Proliferating cell indices in liver lesions. The fraction of proliferating cells (%) from each lesion was calculated from the number of Ki‐67‐positive and total nuclei. The mean rates with SD are shown. Data were analyzed statistically using Tukey’s test. *P < 0.05. (f) Proliferating cell indices in transplanted tumor lines. The mean rates with SD of four independent tumors are shown. Statistical analysis was by Student’s t‐test. *P < 0.05.

Roles of Wnt signal activation in cell cycle regulation.  To investigate further what increases proliferating cells downstream of Wnt signaling, we analyzed the expression profile of cell cycle‐regulating proteins in transplanted HCC (Fig. 6a). Interestingly, expression of cyclin E was upregulated in C1 and C2 compared with wild‐type liver, and in L1 to a lesser extent, whereas the levels of cyclin A, cyclin D1, and c‐Myc in C1 and C2 remained at similar levels to those in L1 and L2 except for robust cyclin D1 expression in L2. An inhibitor of cyclin E/cyclin dependent kinase (CDK) 2 complex, p27 was downregulated. To analyze cyclin E expression in vivo, we immunostained the protein in the primary HCC of Lkb1 ^+/− and compound mutant mice. HCC in the compound mutant mice had significantly high levels of cyclin E, whereas those in Lkb1 ^+/− mice rarely contained positively stained nuclei (Fig. 6b). However, a semiquantitative RT‐PCR analysis showed that the Ccne1 mRNA level was only marginally higher in the HCC of the compound mutant mice than those in Lkb1 ^+/− mice (Fig. 6c). These results suggested that accumulation of cyclin E was mediated by post‐transcriptional regulation.

Figure 6.

 Effects of Wnt signaling on cell cycle regulation. (a) Western blotting analysis of cell cycle‐related proteins. Expression is also shown for β‐catenin and LKB1 to confirm the genotypes of transplanted hepatocellular carcinomas (HCC). (b) Immunostaining for cyclin E in primary HCC of a Lkb1 ^+/− (left) and compound mutant (right) mouse. Scale bars = 100 μm. (c) RT‐PCR analysis for Ccne1 in transplanted (left) and primary (right) HCC. Wild‐type livers (W), transplanted HCC lines (L1, L2, C1, and C2), and pairs of the primary tumor (T) and adjacent normal tissue (N) were analyzed. Expression is shown for Gapdh mRNA as an internal control (lower panels).

Discussion

In the present study, we have developed a new mouse model for HCC by compound mutations in Lkb1 and β‐catenin (Catnb ^+/lox(ex3) Lkb1 ^+/−). We have shown that cells in NdFc of the Lkb1 ^+/− mice lack LKB1 expression because of LOH (Fig. 1), suggesting that loss of LKB1 can be responsible for early steps of hepatocarcinogenesis. Considering the low incidence and multiplicity, and long latency to develop HCC in Lkb1 ^+/− mice, additional genetic alterations are likely to be needed for the progression of NdFc to fully malignant HCC. In this new mouse model of HCC, multiple HCC developed when their livers were induced to express active β‐catenin in Lkb1 ^+/− mice at ∼1 year of age, but not when induced at 2 months of age. Development of dysplastic NdFc in Lkb1 ^+/− mice requires a long latency period (>30 weeks), and no foci can be found in these mice at 2 months of age. It is therefore plausible that HCC in the compound mutant mice arose as a result of β‐catenin activation in a fraction of the dysplastic cells in the pre‐formed NdFc. Thus, we propose a model where NdFc are initiated by the loss of Lkb1, through LOH, and a subset of such NdFc cells undergo activation of β‐catenin, upon AdCMV‐cre infection, and these cells progress further to form HCC.

HCC in the compound mutant mice contained a significantly larger number of proliferating cells than those in Lkb1 ^+/− mice (20 vs 9.8% on average; Fig. 5e), which can explain the accelerated development of HCC in the compound mutant mice (Fig. 2c). In fact, these HCC with the activation of β‐catenin rapidly grew when transplanted into immunocompromised mice (Fig. 2e). In the liver, Wnt signaling plays important roles in zonal expression of liver‐specific genes along the portal‐to‐central axis,^{( 30 )} including those for maturation of normal post‐mitotic hepatocytes. On the other hand, we previously reported that co‐expression of activated β‐catenin and activated H‐ras (H‐rasG12V) in the hepatocytes led to HCC in mice at a high incidence, and that the activation of β‐catenin in this model provoked proliferation and expansion of dysplastic lesions induced by the activated H‐ras.^{( 31 )} Growth advantage of HCC with activating mutations of β‐catenin has also been demonstrated in other studies using human clinical samples or transgenic mouse models.^{( 32} , ³³ , ^{34 )} Accordingly, it is suggested that the Wnt signal activation increases the proliferation rate of the pre‐formed dysplastic hepatocytes induced by other initiating events.

We have shown here that many of the genes known to be induced by Wnt signaling are upregulated in the compound mouse HCC. We found that cyclin E expression was upregulated in the compound mouse HCC, although expression levels remained unchanged in the compound mutant mice for c‐Myc and cyclin D1, well‐known direct targets of Wnt signaling.^{( 35} , ³⁶ , ^{37 )} Cyclin E promotes G₀/S transition through CDK‐dependent and ‐independent mechanisms,^{( 38} , ^{39 )} and ectopic cyclin E induction causes irregular entry of quiescent hepatocytes into the cell cycle.^{( 40 )} It is therefore likely that cyclin E upregulation by Wnt signal activation promoted proliferation of HCC cells in the compound mutant mice, although the precise molecular mechanism of how Wnt signaling leads to cyclin E accumulation remains to be elucidated.

In conclusion, we have constructed a novel mouse model for multistep hepatocarcinogenesis by inducing two sequential genetic changes, where loss of LKB1 in hepatocytes triggers formation of NdFc, with the ensuing activation of β‐catenin leading to the progression of NdFc into HCC. These model mice and their transplantable HCC should help elucidate the molecular mechanism of HCC progression at each step. We also believe that this model represents a subset of human HCC, where LKB1 and β‐catenin mutations are involved (Fig. S4). Further clinical studies will estimate the contributions of the loss of LKB1 and activation to β‐catenin on their pathogenesis and prognosis.

Disclosure Statement

None.

Supporting information

Fig. S1. Immunostainings for glutamine sysnthetase (GS) in (a) wild type and (b) Catnb^lox(ex3) mouse livers. To determine the efficiency of adenoviral infection, we injected 2 × 10⁷ pfu of AdCMV‐cre into Catnb^lox(ex3) mice and necropsied one week later. The activation of β‐catenin was determined by an immunostaining for glutamine synthetase (GS), which is a target gene in hepatocytes.⁽²⁶⁾ Arrows point GS‐positive hepatocytes induced by AdCMV‐cre infection. Arrowheads indicate expression of the endogenous GS around central veins (∗). P, portal vein. Scale bars, 200 μm.

Fig. S2. Levels of up‐regulated genes by activated β‐catenin. RT‐PCR results for representative genes are shown. Wild‐type liver and transplanted HCC lines (L1, L2, C1, and C2) were analyzed.

Fig. S3. Expression of glutamine synthetase (GS) in a wild‐type liver (a), a nodular focus (NdFc) in a Catnb^lox(ex3) mouse (b), an HCC (T) in an Lkb1 ^+/− mouse (c), and an HCC (T) in a compound mutant mouse (d). ∗, central vein. P, portal vein. Scale bars, 500 μm.

Fig. S4. Expression of β‐catenin and LKB1 in human HCCs. Western blotting analysis for β‐catenin and LKB1 (arrowhead) with GAPDH control is shown. Representative 5 pairs of HCC (T) and adjacent tissue (N) are shown. Note β‐catenin accumulation in patient 1, and its truncation in patient 3, accompanied by LKB1 down‐regulation.

Table S1. List of Primer sequences for RT‐PCR.

Tables S2 and S3. Up‐ and down‐regulated genes induced by activated β‐catenin. The changes in the mRNA levels between HCC transplants from the compound mutant mice (C1 and C2) and Lkb1+/− mice (L1 and L2) were analyzed by DNA microarrays. The data are expressed as fold‐changes (up‐regulated: C1/L1 and C2/L2, down‐regulated: L1/C1 and L2/C2).

Please note: Wiley‐Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item