Oncogene (2006) 25, 3528–3536
& 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc
ORIGINAL ARTICLE
Hepatocellular carcinoma in Txnip-deficient mice SS Sheth1, JS Bodnar2, A Ghazalpour1, CK Thipphavong1, S Tsutsumi3, AD Tward4, P Demant5, T Kodama6, H Aburatani3 and AJ Lusis1,7 1
Departments of Human Genetics, Medicine, Molecular Biology Institute, and Microbiology, Immunology, and Molecular Genetics, University of California Los Angeles, 47-123 CHS, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA; 2Beckman Coulter, Inc., Biomedical Research Division, Fullerton, CA, USA; 3Genome Science Division, Research Center for Advanced Science and Technology, University of Tokyo, Meguro-ku, Tokyo, Japan; 4Department of Microbiology and Immunology, University of California San Francisco, GW Hooper Foundation, San Francisco, CA, USA; 5Department of Molecular and Cellular Biology, Roswell Park Cancer Institute, Elm and Carlton Streets, Buffalo, NY, USA; 6Laboratory for Systems Biology and Medicine, Research Center for Advanced Science and Technology 34, University of Tokyo, Meguro-ku, Tokyo, Japan and 7UCLA Johnson Comprehensive Cancer Center, Los Angeles, CA, USA
The molecular pathogenesis and the genetic aberrations that lead to the progression of hepatocellular carcinoma (HCC) are largely unknown. Here, we demonstrate that the thioredoxin interacting protein (Txnip) gene is a candidate tumor suppressor gene in vivo. We previously showed that the recombinant inbred congenic strain HcB19 has a spontaneous mutation of the Txnip gene, and we now show that the strain has dramatically increased incidence of HCC, and that the HCC cosegregates with the Txnip mutation. Approximately 40% of the Txnipdeficient mice developed hepatic tumors with an increased prevalence in male mice. Visible tumors develop as early as 8 months of age. Histological analysis confirmed the morphology of HCC in the Txnip-deficient mice. Molecular markers of HCC, a-fetoprotein and p53, were increased in tumors of Txnip-deficient mice. The upregulation of p53 preceded tumor development; however, bromodeoxyuridine (BrdU) labeling of normal hepatic tissue of Txnip-deficient mice did not reveal increased cell proliferation. Finally, microarray analyses of tumor, nontumor adjacent, and normal tissue of Txnip-deficient mice highlighted the genetic differences leading to the predisposition and onset of HCC. Our findings suggest that Txnip deficiency is sufficient to initiate HCC and suggest novel mechanisms in hepatocarcinogenesis. Oncogene (2006) 25, 3528–3536. doi:10.1038/sj.onc.1209394; published online 10 April 2006 Keywords: hepatocellular carcinoma; mouse model; Txnip; thioredoxin
Correspondence: Professor AJ Lusis, David Geffen School of Medicine at UCLA, 47-123 Center for Health Sciences, Los Angeles, CA 90095-1679, USA. E-mail:
[email protected] Received 19 September 2005; revised 4 November 2005; accepted 7 November 2005; published online 10 April 2006
Introduction Hepatocellular carcinoma (HCC), a cancer of primary hepatocytes, is the most prevalent type of primary tumor of the liver. Although the molecular pathogenesis of HCC is poorly understood, recent expression profiling studies of HCC have led to the identification of candidate genes involved in tumor progression (TackelsHorne et al., 2001; Chen et al., 2002). Mouse models for HCC have provided some insight into the mechanism of development and progression of liver cancer from spontaneous and drug-induced genetic mutations. C3H, an inbred strain susceptible to hepatic tumors, has been studied extensively for spontaneous liver tumors (Maronpot et al., 1995). As part of a study of variations in lipid metabolism among different strains of mice, we have identified a recombinant congenic strain, with a background consisting of 88% C3H/DiSnA and 12% C57BL/10ScSnA, to have an increased incidence of HCC. Fine mapping showed mouse chromosome 3 to segregate with the combined hyperlipidemia, and through a positional cloning approach, the thioredoxin interacting protein gene (Txnip) was identified as underlying for the combined hyperlipidemia phenotype in mice (Pajukanta et al., 2001; Bodnar et al., 2002). We now show that the Txnip deficiency in mice also leads to spontaneous hepatic carcinoma phenotype. Txnip was originally identified as a gene upregulated by 1,25-dihydroxyvitamin D3 in HL-60 cells and, therefore, named vitamin D3 upregulated protein (Vdup1) (Chen and DeLuca, 1994). The Txnip mutation influences aspects of lipid metabolism, glucose metabolism, and also the NADH/NAD þ ratios (Sheth et al., 2005). In addition, Txnip-deficient mice lack functional NK cells (Lee et al., 2005). The major function of Txnip is believed to be inhibition of the redox protein thioredoxin, although other molecular functions of Txnip such as transcriptional regulation have been suggested (Han et al., 2003). There is a growing body of evidence that Txnip may play a causal role in tumor biology. Txnip expression is downregulated in breast and gastric cancers, as well as
Txnip-deficient mice exhibit HCC SS Sheth et al
3529
in many cell lines (Yang et al., 1998; Ikarashi et al., 2002; Takahashi et al., 2002; Escrich et al., 2004). In vitro studies indicate that Txnip overexpression can inhibit the proliferation of stomach cancer and promyelocytic leukemia cells (Han et al., 2003). In a tumor transplant model, Txnip overexpression suppressed tumor growth and metastasis (Goldberg et al., 2003). Recently, it was shown that Txnip interacts with and inhibits JAB1, a protein that promotes the degradation of p27(kip1), a cyclin-dependent kinase inhibitor whose expression is reduced in many tumors (Jeon et al., 2005). However, the question of whether Txnip deficiency can contribute to the initiation and progression of cancer has not been addressed. We now provide evidence that the Txnip deficiency leads to HCC in mice. Expression profiling showed that genetic patterns are consistent between human HCC and the HCC in Txnip-deficient mice. Also, molecular markers, such as a-fetoprotein and p53, which are commonly increased in human HCC, are increased in Txnip-deficient mice. These studies provide the first conclusive in vivo evidence that the loss of Txnip contributes to the development and progression of malignancy, and in particular HCC.
Results Txnip-deficient mice have an increased incidence of HCC As early as 8 months of age, Txnip-deficient mice exhibited hepatic tumors, while no tumors were found in the control strain. None of the 108 control mice that were examined exhibited any evidence of hepatic tumors (Figure 1). However, 47 out of 115 (41%) age and sexmatched Txnip-deficient mice examined had grossly observable hepatic tumors, in some cases involving more than one lobe of the organ (Po0.0001) (Figure 1a). Tumor incidence increased as mice aged. At 8 months of age, Txnip-deficient mice exhibited tumors only rarely (one out of 115 mice). By 13–14 months of age, B19% of the mice had visible tumors, and by 20–24 months B66% had tumors. To determine whether the Txnip deficiency was responsible for the tumor formation, a backcross of the Txnip-deficient mice to strain CAST/Ei was constructed ((HcB-19/Dem CAST/Ei) F1 HcB19/Dem). Of the 180 recombinant backcross progeny analyzed, tumors were seen in 52 out of 108 mice homozygous for the Txnip mutation and four out of 72 mice heterozygous for the mutation. A w2 analysis of these data confirmed that the tumor phenotype segregated with the Txnip mutation (Po108) (data not shown). When comparing the heterozygous mice from the recombinant backcross progeny to the control (C3H/DiSnA) strain, the data demonstrated that mice heterozygous for the Txnip mutation have an increased incidence of tumors (Po0.05). Although the HcB-19 strain has regions of C57BL/10ScSnA, the Txnip mutation arose spontaneously on the C3H/DiSnA background. Based on the genotypes of the microsatellite markers that flank the Txnip gene, D3Mit76 and
Figure 1 Txnip-deficient mice have an increased incidence of grossly obvious hepatic tumors (a). C3H/DiSnA were the control mice, indicated by shaded bars, and Txnip-deficient (Txnip/) mice are indicated by solid bars. Percentage of mice with grossly obvious tumors. (b) Female indicated by shaded male indicated by solid. Percentage of male and female Txnip-deficient mice with grossly obvious tumors. *Represents Po0.05, **represent Po0.01, and ***represent Po0.001. n ¼ number of mice.
D3Mit101, there were no recombination events observed between the two markers in the mice with tumors, indicating perfect cosegregation between tumors and Txnip deficiency. Consistent with HCC in humans, there was also a significant increase in the incidence of tumors in male Txnip-deficient mice in comparison to females. In humans, males are at greater risk of HCC by 3–10 times in comparison to female subjects. In Txnipdeficient mice, the incidence of tumors was 2.4-fold higher in males than in females (Po0.005) (Figure 1b). We conclude that the mutation in Txnip was responsible for the increased tumor incidence. Histological analysis of tumors The histological evidence of the tumors in Txnipdeficient mice confirmed the presence of HCC. Hematoxylin and eosin staining of tumors from Txnipdeficient mice showed cellular architecture characteristic of HCC, with a paucity of bile ducts, absence of lobular architecture, and widened cell plates greater than three cells thick (Figure 2). The histological pattern of the tumors was either trabecular (Figure 2a) or sheet-like (Figure 2c, upper left). As with HCC, the tumors of Txnip-deficient mice were often multifocal. Grossly, some tumors were red to black in color, reflecting hemorrhaging and some displayed fibrosis and the formation of regenerative nodules, indicative of cirrhosis, both of which are commonly present in HCC (data not shown). Other tumors were yellow in color, Oncogene
Txnip-deficient mice exhibit HCC SS Sheth et al
3530
reflecting the presence of fat deposits, a marker for HCC, often in tumor cytoplasm. The fat deposits were also consistent with the fatty liver phenotype observed in the Txnip-deficient mice (Sheth et al., 2005). Macrovesicular fat was also frequently present in Txnipdeficient tumors (Figure 2a).
Figure 2 Immunohistochemistry confirms the presence of hepatocellular carcinoma in Txnip-deficient mice. HCC and normal liver tissue samples of Txnip-deficient mice were sectioned and stained with hematoxylin and eosin (H&E). (a) H&E staining of a representative section of HCC in Txnip-deficient mice at 10 (Inset: 40). (b) H&E staining of a representative section of normal liver tissue of Txnip-deficient mice at 10 (Inset: 40). (c) H&E staining of a normal (indicated as ‘Normal’ in the image) and cancerous (indicated as ‘HCC’ in the image) liver section from a Txnip-deficient mouse at 10.
To further validate the identity of the tumors as HCC, we analysed molecular markers of human HCC, including a-fetoprotein, a fetal glycoprotein, and p53, a tumor suppressor gene (Hsu et al., 1993; Chen et al., 2002). Mutations in p53 also lead to increased p53 protein levels in HCC (Hsu et al., 1993; Chen et al., 2002). Western blot analysis indicated that these markers showed increased expression in the livers of Txnip-deficient mice (Figure 3a, b). Preneoplastic changes in Txnip-deficient livers To gain insight into the mechanism by which Txnip deficiency induces HCC, we analysed livers of Txnipdeficient mice prior to tumor formation. We labeled cells from normal, young Txnip-deficient mice (3 months of age) with 5-bromo-20 -deoxyuridine (BrdU) to determine whether there was increased cell proliferation or dysplasia prior to tumor onset. There was no obvious difference in the number of proliferating cells in the Txnip-deficient mice in comparison to control mice (n ¼ 3 in each group) (Figure 3c, d). The labeling index (LI), or the number of proliferative cells as a percentage of the total number of cells, was 2.571.4 in Txnipdeficient mice versus an LI of 2.671.7 in the control mice (data not shown). Hence, deficiency of Txnip alone is not sufficient to induce proliferation in hepatocytes. To determine if the hepatocytes of Txnip-deficient mice are experiencing a genotoxic stress, we examined p53 as a surrogate marker (Bartkova et al., 2005). We observed increased levels of p53 in the pretumorous livers of Txnip-deficient mice (3 months of age), but levels of a-fetoprotein in the normal livers of Txnip-deficient mice were similar to the control mice (Figure 3a, b). Hence, even prior to the onset of tumor,
Figure 3 Txnip-deficient HCC expresses increased levels of HCC markers without increased cell proliferation prior to tumor onset. Western blot analysis of liver protein samples and BrdU-labeled cells from Txnip-deficient mice and control mice (C3H/DiSnA). (a) Liver a-fetoprotein (afp) levels in Txnip-deficient mice with tumors, control livers, and normal Txnip-deficient mice. (b) Liver p53 levels in Txnip-deficient mice with tumors, control livers, and normal Txnip-deficient mice. (c) BrdU-labeled, Txnip-deficient hepatocytes at 20 (Inset: BrdU-labeled, Txnip-deficient intestine as a staining control). (d) BrdU-labeled control (C3H/DiSnA) hepatocytes at 20 (inset: BrdU-labeled control intestine as a staining control). Oncogene
Txnip-deficient mice exhibit HCC SS Sheth et al
3531
Txnip-deficient mice are experiencing some p53-inducing stimulus, perhaps DNA damage. Vitamin D regulates Txnip in vivo Txnip was originally identified as a protein that is upregulated by 1,25-dihydroxyvitamin D3 in vitro (Chen and DeLuca, 1994). We therefore investigated whether Txnip is upregulated by 1,25-dihydroxyvitamin D3 in vivo. The C3H/DiSnA control strain was fed either a diet deficient in 1,25-dihydroxyvitamin D3 or the appropriate control diet for 36 weeks. We observed a significant decrease in the mRNA levels in the livers of mice fed the 1,25-dihydroxyvitamin D3-deficient diet as compared to the control diet (Figure 4). We then dissected the mice at 12–14 months of age to determine whether the lack of 1,25-dihydroxyvitamin D3 would affect the incidence of hepatic tumors. None of the 30 mice analysed showed evidence of hepatic tumors based on gross morphology. Unfortunately, multiple efforts to raise antibodies to Txnip in our laboratory have failed, and therefore we have been unable to determine protein levels in these or other experiments. Microarray data patterns in Txnip-deficient mice We used Affymetrix expression array chips (Mouse Genome 430A 2.0. array) to compare the transcript profiles of tumor and normal liver tissues of Txnipdeficient mice. We profiled the tumor and non-tumor, adjacent tissue obtained from the same mouse to account for differences caused by the paracrine effect of tumor cells on normal, adjacent cells in Txnipdeficient mice. This experiment was comparable to human studies in which the transcription profile of the tumor sample was referenced against the non-tumor tissue from the same patient. The liver samples from Txnip-deficient mice were subdivided into five groups: tumor samples from mice at 22 months of age, 24 months of age, and 27 months of age, along with normal samples from mice at 22 months of age and non-tumor, adjacent samples from the mice with tumors that were 24 months of age. The results indicated the presence of B1090 genes with a differential gene expression of at
Figure 4 1,25 dihydroxyvitamin D3 upregulates the Txnip gene in vivo. Northern blot analysis of Txnip RNA in C3H/DiSnA (control) mice. Control mice were fed either a 1,25 dihydroxyvitamin D3-deficient diet (Vit. D Deficient Diet) or the appropriate control diet (Control Diet). Txnip RNA levels were measured in the liver.
least twofold among the five groups of tissues profiled. We performed partitioning around medoids (PAM) clustering of the 1090 genes, which subdivided the genes into six groups based on the gene expression profile similarities (Dudoit and Fridlyand, 2002). Each PAM, indicated by PAM1, PAM2, PAM3, PAM4, PAM5, and PAM6, contained 81, 186, 295, 330, 134, and 64 genes, respectively (for the complete list of genes and accession numbers, see Supplementary Tables 1–6). In the first four PAM groups (PAM1, PAM2, PAM3, and PAM4), tumor and non-tumor samples were separated into two distinct clusters (Figure 5a–d). The first three PAMs (PAM1–3) contained genes that were overexpressed in tumor tissues in comparison to the normal samples, with PAM1 representing modest differences and PAM3 representing the most highly differentially expressed genes in the tumor samples. PAM4 contained genes that were downregulated in tumor samples. In PAM5, all samples showed similar expression profiles, except for the tumor samples from the 27-month-old Txnipdeficient mice, suggesting a possible age-related difference (Figure 5e). Lastly, in PAM6, all samples showed a similar expression profile, except for non-tumor, adjacent tissues obtained from mice with tumors (Figure 5f). This cluster may represent genes in non-transformed hepatocytes that were subject to the stress induced by the adjacent tumor cells. We also performed hierarchical clustering on the 1090 genes to identify the overall variations in transcript levels. The pattern of gene expression for these genes resulted in the expected clustering of samples into two distinct groups: normal hepatic tissue and tumor tissue (data not shown). The list of the genes in each PAM contains genes that have been previously associated with HCC and/or cancer in general. In PAM3, which contains highly differentially expressed genes in the tumor samples, Tnfrsf12a, an immediate early response gene induced by serum and growth factors in cell culture, ranks among the top genes (>20-fold). Also, endothelial cell-specific molecule-1 (Esm1), which effects mitogenic activity, is another gene found to be highly upregulated in the tumors of Txnip-deficient mice (>30-fold). Interestingly, E-Cadherin, a marker for HCC, which has been shown to be directly associated with HCC, was increased 10-fold in the Txnip-deficient tumors (Jiao et al., 2002; Wei et al., 2002). b-Catenin is a marker for HCC and an accumulation of it correlates with tumor size (Buendia, 2002). In the tumor tissue, there was a 1.5-fold increase in comparison to the normal hepatic tissue of Txnip-deficient mice. The tumor samples from Txnip-deficient mice also had increased expression of genes regulating the cell cycle. DNA replication genes, Mcm2, 3, and 6, proto-oncogenes, Nmyc1 and c-Jun, and cell cycle regulation genes, Cdkn2b and Ccnb1, were some of genes upregulated in Txnip-deficient mice. PAI1, which is upregulated in hepatic cancer cells, was increased 10-fold in the tumors (Grondahl-Hansen et al., 1993; Zhou et al., 2000). DDr1, a gene found to be upregulated in ovarian cancer, was increased in the tumor samples by fivefold (Heinzelmann-Schwarz et al., 2004). Lastly, we observed a >10-fold increase in Tff3, Oncogene
Txnip-deficient mice exhibit HCC SS Sheth et al
3532
Figure 5 In Txnip-deficient mice, partitioning around medoids (PAM) clustering of 1090 differentially expressed genes under five conditions are divided into six clusters. The five different conditions of the Txnip-deficient mice profiled are: tumor (22 months), tumor (24 months), tumor (27 months), non-tumor, adjacent (24 months), and normal. The 24-month-old Txnip-deficient mice had portions of the liver that were cancerous and normal (indicated as tumor or normal). (a) PAM1; (b) PAM2; (c) PAM3; (d) PAM4; (e) PAM5. (f) PAM6. Each PAM represents a cluster showing expression differences between conditions. Each figure shows the heat map generated from the expression levels of the genes present in the cluster with some of the cancer-related genes highlighted on the right and mentioned in the text. The corresponding dendogram on the left of each map shows the degree of similarity between samples (as indicated as 1, 2, 3, 4, 5 on the heat map) for each PAM.
Oncogene
Txnip-deficient mice exhibit HCC SS Sheth et al
3533
another gene that has been associated with a wide variety of cancers (Faith et al., 2004; Solmi et al., 2004). Some novel genes that were highly upregulated in the Txnip-deficient tumors, but not associated with cancer previously, are: Rex3, Nope, Ly6d, Tpm1, Slpi, Scd2, Fabp5, Fabp4, Mep1b, and Xmr. Approximately one-third of the differentially expressed genes were the genes that were downregulated in tumor tissues. Among the downregulated genes were liver-specific genes, major urinary protein (MUP), and members of the cytochrome p450 family. Other downregulated genes included inhibitors of nucleotide signaling (Gdi2), immune system, and host defense response genes (CD3, Fcgr2b), and a form of super-oxide dismutase (SOD3). Bat, a potential marker for the prognosis of HCC, was also decreased in tumor samples (Furutani et al., 1996). The sixth cluster (PAM6) contained genes that were downregulated in non-tumor, adjacent tissues of Txnip-deficient mice (Figure 5). Among the 69 genes in this cluster, genes involved in cell cycle checkpoints (Wee1), DNA synthesis and repair (RR), and MAP-kinase signaling proteins (ELK4, MAPkinase interacting protein 2, Dusp7) were significantly downregulated. Also, immune system genes (Igh-4, Il1r2, immunoglobulin kappa chain variable 28), ubiquitin-specific proteases (Usp2, Usp9x, Usp12), and genes involved in transcription regulation (Hdac7a, Creb1) showed a significant decrease in expression in the nontumor, adjacent tissue. The role of these genes in the paracrine effect of tumor cells on gene transcription regulation program remains to be elucidated.
Discussion There are a number of potential mechanisms by which Txnip deficiency may cause HCC in mice. One possibility is that loss of Txnip causes non-specific damage to the liver that results in multiple rounds of replication that create a permissive environment for mutations to arise. This possibility is unlikely as pretumorous livers from Txnip-deficient mice did not show increased proliferation, nor did they exhibit histological changes characteristic of damaged liver (Figure 2b). Another possibility is that Txnip may have a direct role in restricting cell cycle progression. Previous in vitro studies of cell lines from other tumor types lend support to this hypothesis (Han et al., 2003). Although we observed no increase in proliferation in the pretumorous liver, Txnip deficiency may act in concert with other proliferative stimuli to induce tumor progression. This possibility is also supported by Txnip’s ability to inhibit the activity of thioredoxin, which is known to promote cell cycle progression (Powis and Montfort, 2001). Txnip has also been shown to interact with transcription factors and chromatin remodeling complexes which may play a role in cell cycle progression (Han et al., 2003). Recently, Txnip-deficient fibroblasts were shown to exhibit increased proliferation, with reduced expression
of p27(kip1), a cyclin-dependent kinase inhibitor. This appears to involve the interaction of Txnip with JAB1, blocking the JAB1-mediated translocation of p27(kip1) from the nucleus to the cytoplasm (Jeon et al., 2005). Txnip may also have a role in inducing apoptosis in response to some stimuli. Minn et al. recently demonstrated that the overexpression of human TXNIP induced apoptosis in pancreatic b cells as determined by Bax, Bcl2, caspase-3, and cleaved caspase-9 expression (Minn et al., 2005). It has also been shown that Txnip mRNA expression was upregulated in during neuronal apoptosis (Saitoh et al., 2001). Thioredoxin has been shown to have antiapoptotic properties, possibly mediated through ASK1 binding or antioxidant properties. Hence, loss of Txnip may accentuate these properties of thioredoxin as well. An additional mechanism by which Txnip deficiency may be promoting the development of HCC is through alterations of redox status. Txnip has been shown to affect the redox status as measured by NADH and NAD þ ratios (Sheth et al., 2005). CuZnSOD-deficient mice show disruptions in redox homeostasis as well as an increase in HCC incidence (Elchuri et al., 2005). Also, ethanol treatment of mice has been shown to inhibit liver regeneration and induce redox sensitive cell cycle inhibitors (Koteish et al., 2002). As indicated above, thioredoxin has multiple roles which are regulated in a redox sensitive fashion (Powis and Montfort, 2001). There may be a possibility of Txnip functioning similarly to the HBV-X protein (HBx). HBx has been shown to affect apoptosis, signal transduction pathways, and oxidative stress through direct interactions with mitochondrial proteins (Lee et al., 2004). Upon the expression of HBx, there are increased levels of cellular ROS in hepatoma cell lines (Lee et al., 2004). Studies have also implicated HBx in the onset of HCC (Kim et al., 1991; Feitelson and Duan, 1997). It has been suggested that the oxidative stress, potentially as a direct result of HBx or as a result of the host’s immune system, is the cause of liver diseases, such as HCC. Based on the evidence that HBx modulates mitochondrial enzymes, increases ROS production, and may lead to the onset of HCC, it may be possible that the increased oxidative stress due to HBx is leading to the activation of thioredoxin and thereby associated to the HCC observed in the Txnip-deficient mouse. The expression profiling of the tumor and normal tissues of the Txnip-deficient mice revealed altered expression of genes involved in a broad range of molecular functions, and involved in various types of cancer. Each of the PAM clusters suggested differences between the tumor and normal tissues in cell cycle regulation, proliferation, apoptosis, cell adhesion, cytoskeleton rearrangements, DNA binding, transcription, lipid binding, lipoprotein metabolism, and signaling. Of particular interest were potential markers for HCC, such as Tnfrsf12a and Esm1. In comparison to normal hepatic tissue, Tnfrsf12a is elevated in murine models for HCC and in tumor specimens from HCC patients (Feng et al., 2000). Esm1 binds to hepatocyte growth Oncogene
Txnip-deficient mice exhibit HCC SS Sheth et al
3534
factor and increases its mitogenic activity, and overexpression of Esm1 has been shown to induce tumor growth in SCID mice (Bechard et al., 2001; Trowbridge and Gallo, 2002; Scherpereel et al., 2003). Also, PAI-1, which was significantly increased in the tumors of the Txnip-deficient mice, is a gene that has been associated with poor prognosis in cancer patients (GrondahlHansen et al., 1993; Zhou et al., 2000). We identified genes that were highly differentially expressed in the tumor samples and have also been implicated in various types of cancer, such as DDr1 and Tff3. Lastly, the downregulation of liver-specific genes, MUP and CYP450, in the tumors was consistent with previous microarray studies suggesting that dedifferentiation of hepatocytes is a step toward the transformation of hepatocytes (Dragani et al., 1989; Yamada et al., 1999). We have now shown that Txnip is regulated by 1,25dihydroxyvitamin D3 in vivo. Inhibitory effects of 1,25dihydroxyvitamin D3 on the proliferation of a variety of cancer cell lines have been reported (Capiati et al., 2004; Chen et al., 2004; Jiang et al., 2004; Sergeev, 2004). 1,25dihydroxyvitamin D3 has been shown to inhibit proliferation of human HCC cell lines that express vitamin D receptor; however, 1,25-dihydroxyvitamin D3 does not show antiproliferative effects on cell lines that do not express the receptor (Miyaguchi and Watanabe, 2000). Chronic treatment of a hepatoblastoma cell line, HepG2, with 1,25-dihydroxyvitamin D3 leads to the inhibition of cell proliferation (Pourgholami et al., 2000). In summary, we provide the first in vivo evidence that Txnip deficiency is sufficient to induce tumor formation. These mice provide a valuable model for studying the as yet poorly characterized link between redox signaling and cancer. Based on our results, genetic studies to test for loss of Txnip in human tumors appear warranted.
Materials and methods Animal husbandry and diets The recombinant congenic strain, HcB-19/Dem (Txnip deficient), was developed as described previously (Demant and Hart, 1986; Castellani et al., 1998; Bodnar et al., 2002). The Txnip mutation arose on the HcB-19/Dem, which has a genetic background of strain, C3H/DiSnA. Therefore, C3H/DiSnA served as the parental control strain for the experiments where applicable. All mice were housed under conditions meeting the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care and were housed in groups of five or less animals per cage and maintained on a 12 h light– dark cycle at an ambient temperature of 231C. They were allowed ad libitum access to water and rodent chow containing 6% fat (Diet No. 8604, Harlan Teklad). In the 1,25dihydroxyvitamin D3 deficiency analysis, the control mice were on a 1,25-dihydroxyvitamin D3-deficient diet (TD 89123, Harlan Teklad) containing 0.47% calcium and 0.3% phosphate or the appropriate 1,25-dihydroxyvitamin D3 control diet (TD 89124, Harlan Teklad) for 36 weeks. All animal care and experimental protocols were approved by the UCLA Animal Research Committee. Oncogene
Immunohistochemistry Txnip-deficient and control mice were anesthetized by isofluorane and killed by cervical dislocation. Tissue samples were fixed in 10% formalin, embedded in paraffin, sectioned, stained with hematoxylin–eosin, and examined by using conventional optical microscopy. Bromodeoxyuridine (BrdU) labeling DNA synthesis in Txnip-deficient and control mice was determined by measuring the incorporation of BrdU into DNA using a cell proliferation kit (Cat. #: 00–0103; Zymed Laboratories). The amount injected into each mouse was 1 ml/ 100 g body weight. The mice were killed after 6 h. The liver was removed along with the small intestine as a staining control. Tissue samples were fixed in 10% formalin, embedded in paraffin, sectioned, and stained for BrdU according to the manufacturer’s instructions (Cat. #: 93–3943; Zymed Laboratories). BrdU labeling of hepatocytes was determined under light microscope ( 100) and expressed as the labeling index (LI), which is a percentage of all hepatocytes that were positive. A total of three mice were there in each group (with four areas per liver lobe examined). Western blot analyses Mice were anesthetized with isofluorane and killed by cervical dislocation. Liver tissue was immediately placed in liquid nitrogen, which was weighed and added to a lysis buffer (20 mM HEPES, 100 mM KCl, 300 mM NaCl, 10 mM EDTA, 0.1% Nonidet P-40, and a cocktail of protease inhibitors (Roche). Protein samples were treated for Western blotting according to the manufacturer’s instructions (Invitrogen). Membranes were blocked with phosphate-buffered saline with 0.1% Tween 20/5% nonfat dry milk overnight, washed, and incubated with primary and secondary antibodies for 2 and 1 h, respectively. The primary antibodies for a-fetoprotein and p53 (Santa Cruz Biotechnologies) are commercially available antibodies. The secondary antibodies used for a-fetoprotein and p53 were anti-goat IgG (1 : 500) (Calbiochem) and antimouse IgG (1 : 1000) (Amersham), respectively. The membranes were washed and detected by using the ECL Western blotting kit (Amersham) according to the manufacturer’s suggested protocol. Protein levels for a-fetoprotein and p53 were normalized against GAPDH protein levels to ensure equal loading. Northern blot analysis Control mice were anesthetized with isofluorane and killed by cervical dislocation. Total RNA was isolated from liver with TRIzol reagent according to the manufacturer’s instructions (Invitrogen). Total RNA (15 mg) was resolved by formaldehyde/agarose gel electrophoresis, transferred to a nylon membrane, and hybridized with 32P-labeled Txnip cDNA probe as previously described (Shih et al., 2000). Microarray procedure Txnip-deficient mice were killed by cervical dislocation, and the liver tissue was immediately dissected and snap frozen in liquid nitrogen and stored at 801C. Frozen tissue was homogenized in TRIzol (Invitrogen) for total RNA extraction, and the quality of RNA was determined on an Agilent bioanalyser (Agilent Technologies). All tumor, non-tumor adjacent, and normal liver samples were age- and sex-matched with n ¼ 3 in each group ((tumor (22 months), tumor (24 months), tumor (27 months), non-tumor, adjacent (24 months), and normal). The 24-month-old Txnip-deficient mice had portions of the liver that were cancerous and normal
Txnip-deficient mice exhibit HCC SS Sheth et al
3535 (indicated as non-tumor, adjacent). Each sample was arrayed for a total of 15 arrays used for the analysis. The sample preparation protocol from the Affymetrix GeneChips Expression Analysis Manual was used (Affymetrix). Biotin-labeled cRNA was synthesized from 3 to 5 mg of RNA and hybridized to GeneChips Mouse Genome 430A array (Affymetrix). GeneChips Mouse Genome 430A arrays were used to compare tumor, non-tumor adjacent, and normal tissues. After washing, arrays were stained with streptavidin–phycoethrin and analysed on a Hewlett-Packard scanner. Each array was analysed with Microarray Analysis Suite Version 5 to calculate the average density for each probe set on the array (Affymetrix). Average density values were normalized, scaled to 100, and log transformed. Data normalization and analyses were also performed with Microarray Analysis Suite Version 5 software. A comparison of tumor, non-tumor adjacent, and normal tissues was carried out using GeneSifter from which genes with changes in expression levels of twofold or greater were clustered (www.genesifter.net). For hierarchical clustering, GeneSifter was also used. The microarray data discussed in this publication have been deposited in NCBI’s Gene
Expression Omnibus and are accessible through GEO Series Accession Number (GSE:2127). Statistical analyses Data are presented as mean7s.e.m., with n indicating the number of mice. The statistical significance of the presence or absence of tumors between the two groups was determined with ANOVA. Differences were considered statistically significant at Po0.05. We determined the statistical significance of the differentially expressed genes between the following groups of liver samples from Txnip-deficient mice (with n ¼ 3 in each condition): tumor (22 months), tumor (24 months), tumor (27 months), non-tumor adjacent (24 months), and normal using ANOVA and corrected the P-value obtained for multiple comparisons using the Benjamini–Hoescberg method. Acknowledgements This study was supported by NIH Grant HL28481 (AJL) and the Johnson Comprehensive Cancer Center, UCLA.
References Bartkova J, Horejsi Z, Koed K, Kramer A, Tort F, Zieger K et al. (2005). Nature 434: 864–870. Bechard D, Gentina T, Delehedde M, Scherpereel A, Lyon M, Aumercier M et al. (2001). J Biol Chem 276: 48341–48349. Bodnar JS, Chatterjee A, Castellani LW, Ross DA, Ohmen J, Cavalcoli J et al. (2002). Nat Genet 30: 110–116. Buendia MA. (2002). Med Pediatr Oncol 39: 530–535. Capiati DA, Rossi AM, Picotto G, Benassati S, Boland RL. (2004). J Cell Biochem 93: 384–397. Castellani LW, Weinreb A, Bodnar J, Goto AM, Doolittle M, Mehrabian M et al. (1998). Nat Genet 18: 374–377. Chen F, Wang Q, Wang X, Studzinski GP. (2004). Cancer Res 64: 5425–5433. Chen KS, DeLuca HF. (1994). Biochim Biophys Acta 1219: 26–32. Chen X, Cheung ST, So S, Fan ST, Barry C, Higgins J et al. (2002). Mol Biol Cell 13: 1929–1939. Demant P, Hart AA. (1986). Immunogenetics 24: 416–422. Dragani TA, Manenti G, Sacchi MR, Colombo BM, Della Porta G. (1989). Mol Carcinog 2: 355–360. Dudoit S, Fridlyand J. (2002). Genome Biol 3, RESEARCH0036. Elchuri S, Oberley TD, Qi W, Eisenstein RS, Jackson Roberts L, Van Remmen H et al. (2005). Oncogene 24: 367–380. Escrich E, Moral R, Garcia G, Costa I, Sanchez JA, Solanas M. (2004). Mol Carcinog 40: 73–78. Faith DA, Isaacs WB, Morgan JD, Fedor HL, Hicks JL, Mangold LA et al. (2004). Prostate 61: 215–227. Feitelson MA, Duan LX. (1997). Am J Pathol 150: 1141–1157. Feng SL, Guo Y, Factor VM, Thorgeirsson SS, Bell DW, Testa JR et al. (2000). Am J Pathol 156: 1253–1261. Furutani M, Arii S, Higashitsuji H, Mise M, Niwano M, Harada T et al. (1996). Hepatology 24: 1441–1445. Goldberg SF, Miele ME, Hatta N, Takata M, Paquette-Straub C, Freedman LP et al. (2003). Cancer Res 63: 432–440. Grondahl-Hansen J, Christensen IJ, Rosenquist C, Brunner N, Mouridsen HT, Dano K et al. (1993). Cancer Res 53: 2513–2521. Han SH, Jeon JH, Ju HR, Jung U, Kim KY, Yoo HS et al. (2003). Oncogene 22: 4035–4046. Heinzelmann-Schwarz VA, Gardiner-Garden M, Henshall SM, Scurry J, Scolyer RA, Davies MJ et al. (2004). Clin Cancer Res 10: 4427–4436.
Hsu HC, Tseng HJ, Lai PL, Lee PH, Peng SY. (1993). Cancer Res 53: 4691–4694. Ikarashi M, Takahashi Y, Ishii Y, Nagata T, Asai S, Ishikawa K. (2002). Anticancer Res 22: 4045–4048. Jeon JH, Lee KN, Hwang CY, Kwon KS, You KH, Choi I. (2005). Cancer Res 65: 4485–4489. Jiang F, Bao J, Li P, Nicosia SV, Bai W. (2004). J Biol Chem 279: 53213–53221. Jiao W, Miyazaki K, Kitajima Y. (2002). Br J Cancer 86: 98–101. Kim CM, Koike K, Saito I, Miyamura T, Jay G. (1991). Nature 351: 317–320. Koteish A, Yang S, Lin H, Huang J, Diehl AM. (2002). Alcohol Clin Exp Res 26: 1710–1718. Lee KN, Kang HS, Jeon JH, Kim EM, Yoon SR, Song H et al. (2005). Immunity 22: 195–208. Lee YI, Hwang JM, Im JH, Lee YI, Kim NS, Kim DG et al. (2004). J Biol Chem 279: 15460–15471. Maronpot RR, Fox T, Malarkey DE, Goldsworthy TL. (1995). Toxicology 101: 125–156. Minn AH, Hafele C, Shalev A. (2005). Endocrinology 146: 2397–2405. Miyaguchi S, Watanabe T. (2000). Hepatogastroenterology 47: 468–472. Pajukanta P, Bodnar JS, Sallinen R, Chu M, Airaksinen T, Xiao Q et al. (2001). Mamm Genome 12: 238–245. Pourgholami MH, Akhter J, Lu Y, Morris DL. (2000). Cancer Lett 151: 97–102. Powis G, Montfort WR. (2001). Annu Rev Pharmacol Toxicol 41: 261–295. Saitoh T, Tanaka S, Koike T. (2001). J Neurochem 78: 1267–1276. Scherpereel A, Gentina T, Grigoriu B, Senechal S, Janin A, Tsicopoulos A et al. (2003). Cancer Res 63: 6084–6089. Sergeev IN. (2004). J Steroid Biochem Mol Biol 89–90: 419–425. Sheth SS, Castellani LW, Chari S, Wagg C, Thipphavong CK, Bodnar JS et al. (2005). J Lipid Res 46: 123–134. Shih DM, Xia YR, Wang XP, Miller E, Castellani LW, Subbanagounder G et al. (2000). J Biol Chem 275: 17527–17535. Oncogene
Txnip-deficient mice exhibit HCC SS Sheth et al
3536 Solmi R, De Sanctis P, Zucchini C, Ugolini G, Rosati G, Del Governatore M et al. (2004). Int J Oncol 25: 1049–1056. Tackels-Horne D, Goodman MD, Williams AJ, Wilson DJ, Eskandari T, Vogt LM et al. (2001). Cancer 92: 395–405. Takahashi Y, Nagata T, Ishii Y, Ikarashi M, Ishikawa K, Asai S. (2002). Oncol Rep 9: 75–79. Trowbridge JM, Gallo RL. (2002). Glycobiology 12: 117R–125R.
Wei Y, Van Nhieu JT, Prigent S, Srivatanakul P, Tiollais P, Buendia MA. (2002). Hepatology 36: 692–701. Yamada Y, Karasaki H, Matsushima K, Lee GH, Ogawa K. (1999). Lab Invest 79: 1059–1067. Yang X, Young LH, Voigt JM. (1998). Breast Cancer Res Treat 48: 33–44. Zhou L, Hayashi Y, Itoh T, Wang W, Rui J, Itoh H. (2000). Pathol Int 50: 392–397.
Supplementary Information accompanies the paper on Oncogene website (http://www.nature.com/onc).
Oncogene
