Oncogene (2006) 25, 599–608 & 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc ORIGINAL ARTICLE Overexpression of regenerating islet-derived 1 alpha and 3 alpha in human primaryliver tumors with b-catenin mutations

C Cavard1,5, B Terris1,2,5, G Grimber1, L Christa3, V Audard1, B Radenen-Bussiere2, M-T Simon3, C-A Renard4, M-A Buendia4 and C Perret1

1De´partement GDPM, INSERM U-567, CNRS UMR 8104, Institut Cochin, Universite´ Paris 5, Paris, France; 2Service d’Anatomie Pathologique, Hoˆpital Cochin, Universite´ Paris 5, Paris Cedex, France; 3INSERM U-370, Institut Necker-Pasteur, Universite´ Paris 5, Paris Cedex, France and 4Unite d’Oncogene`se et Virologie Mole´culaire, INSERM U-579, Institut Pasteur, Paris Cedex, France

The Wnt/b-catenin signaling pathwayis activated in many although several genetic alterations have been impli- human hepatocellular carcinomas (HCC). We tried to cated in at least three carcinogenesis pathways, p53, RB identifythe genes involved in carcinogenesis and progres- and the Wnt/b-catenin signaling pathway (Buendia, sion of HCC with b-catenin mutations. We used PCR- 2000; Thorgeirsson and Grisham, 2002). Deregulation based subtractive hybridization to compare expres- of the Wnt pathway appears to be the most frequent in sion between malignant and benign components of a human HCC and is detected in about 30–40% of these human HCC occurring in pre-existing adenoma activated patients (de La Coste et al., 1998; Buendia, 2000). for b-catenin. Two of the genes identified belong to the Mutations affecting two partners of the Wnt pathway Regenerating gene (REG) family. They encode the have been found in liver cancers. Activating mutations Regenerating islet-derived 3 alpha (REG3A/HIP/PAP/ in the b-catenin gene occur mainly in HBV-negative REG-III) and 1 alpha (REG1A) , both involved in HCC (Laurent-Puig et al., 2001). The others are liver and pancreatic regeneration and proliferation. Using mutations that inactivate the AXIN 1, and more rarely siRNA directed against b-catenin, we demonstrated that the AXIN 2 gene (Satoh et al., 2000; Laurent-Puig et al., REG3A is a target of b-catenin signaling in Huh7 2001; Taniguchi et al., 2002). hepatoma cells. The upregulation of REG3A and REG1A b-catenin is central to the Wnt pathway. It has a expression is significantlycorrelated to the b-catenin structural role in cell–cell adhesion, but is also a status in 42 HCC and 28 hepatoblastomas characterized transcription cofactor with T-cell factor/lymphoid en- for their b-catenin status. Thus, we report strong evidence hancer factor (TCF/LEF) in the Wnt pathway. In the that both genes are downstream targets of the Wnt absence of Wnt signaling, the level of b-catenin is low. pathwayduring liver tumorigenesis. b-catenin is phosphorylated at critical NH2-terminal Oncogene (2006) 25, 599–608. doi:10.1038/sj.onc.1208860; residues by the glycogen synthase kinase 3-b (GSK3-b) published online 28 November 2005 bound to a scaffolding complex of axin and adenoma- tous polyposis , and subsequently, the phos- Keywords: b-catenin; REG1A; REG3A; hepatocellular phorylated protein is degraded by the ubiquitin– carcinoma; hepatoblastoma; Wnt signaling proteasome system. Wnt stimulation leads to the inactivation of GSK3-b, resulting in the stabilization of b-catenin in the cytoplasm such that it is available to bind the TCF/LEF family of transcription factors and induce target (Giles et al., 2003). The Introduction genetic program triggered by activation of b-catenin signaling depends on the cellular context (Polakis, 2000; Hepatocellular carcinoma (HCC) is the most common Giles et al., 2003). The identification of b-catenin target type of primary . It is the fifth most frequent genes would therefore help understand how aberrant cancer worldwide, and its incidence is still rising in b-catenin signaling contributes to liver carcinogenesis. Western countries (Parkin et al., 2001; El-Serag, 2002). Several studies have investigated the downstream target The prognosis is poor. The molecular changes under- genes of this pathway. Most of these studies involved lying HCC remain largely unknown (Bruix et al., 2004), differential expression analyses using animal models of hepatocarcinogenesis (Cadoret et al., 2002; Ovejero Correspondence: Dr C Cavard, De´ partement GDPM, INSERM U- et al., 2004), human hepatomas cell lines (Yamamoto 567, CNRS UMR 8104, Institut Cochin, Universite´ Paris 5, 24 rue du et al., 2003) and ex vivo infection of primary human Faubourg Saint-Jacques, Paris 75014, France. hepatocytes (Levy et al., 2002), which expressed an E-mail: [email protected] 5These authors contributed equally to this work. oncogenic form of b-catenin. Very few genes specifically Received 2 March 2005; revised 9 May 2005; accepted 19 May 2005; induced in the liver in response to b-catenin have been published online 28 November 2005 identified (Cadoret et al., 2002; Yamamoto et al., 2003; Overexpression of REG genes in human primary liver tumors C Cavard et al 600 Ovejero et al., 2004). Among them, the glutamine- synthetase gene (GS) appears the most reliable marker (Cadoret et al., 2002; Zucman-Rossi et al., submitted). To identify novel candidate b-catenin/TCF transcrip- tional targets likely to contribute to liver carcinogenesis, we used PCR-based subtractive hybridization to com- pare gene expression between the different components of a particular human HCC occurring in pre-existing adenoma activated for b-catenin. We report the identification of two genes that belong to the family of the Regenerating (REG) genes that encode growth- promoting : the Regenerating islet-derived 3 alpha (REG3A) (also named HIP/PAP for Hepatocarcinoma– Intestine–/Pancreatitis Associated-Protein) and the Regenerating islet-derived 1 alpha (REG1A) genes (Zhang et al., 2003). We tested whether the REG3A and REG1A genes are downstream targets of b-catenin and analysed their expression in a panel of primary liver tumors, both HCC and hepatoblastomas (HB), char- acterized for the presence of b-catenin mutations, and also in colon cancers associated with aberrant b-catenin signaling, from familial adenomatous polyposis (FAP) patients. Figure 1 Description of the tumors used for the SSH procedure. (a) Macroscopic view of the HCC (HCC) in pre-existing Results adenoma (Ad). These tumors arose in non-cirrhotic liver (NT). (b) Immunostaining for b-catenin (magnification  200). Intense Cloning of REG3A and REG1A genes from a rare case of nuclear staining was observed in HCC. (c) Immunostaining for GS (magnification  100). Intense cytoplasmic staining is present in HCC occurring in a b-catenin-mutated adenoma the adenoma component. The adjacent non-tumoral liver displayed We studied a particularly rare case of HCC to assess the usual staining around the central vein. (d) RNAs used for the the molecular consequences of a dysregulation of subtracted libraries was analysed for GS mRNA by Northern the Wnt pathway from benign to malignant stages of blotting. NT: non-tumoral; Ad: adenoma; HCC: hepatocellular hepatic tumor development. This male patient displayed carcinoma. The blot was standardized with a human GAPDH probe. an HCC in a pre-existing liver adenoma in a morpho- logically normal liver (Figure 1a). In both components, adenoma and HCC, Wnt/b-catenin signaling was activated as evidenced by the presence of b-catenin specific for the HCC stage as HCC cDNAs have been not only at the membrane but also in the cytoplasm subtracted by sequences expressed in the adenomatous and more infrequently in the nucleus of tumor tissue. cells (Figure 1b). The GS gene is a reliable marker of In all, 78 positive clones corresponding to nine Wnt/b-catenin pathway activation in HCC (Cadoret independent genes were isolated from the adenoma- et al., 2002; Zucman-Rossi et al., submitted), and we enriched library and of 115 positive clones corresponding confirmed activation of the b-catenin pathway by to 27 independent genes from the carcinoma-enriched immunohistochemistry and Northern blot analyses library. The reidentification of the GS gene (gene ID and observed GS overexpression in both stages (Figures 2752) in our suppression subtractive hybridization 1c and d). approach validated the subtraction procedure and served The activation of the Wnt pathway in the adenoma as a positive control. Some of the genes were analysed and HCC was owing to an in-frame 15 base deletion by real-time RT–PCR to confirm their differential involving exon 3 (Dcodons 32–36) eliminating the expression in the tissues from which the RNA had residues that are targets for the destruction of b-catenin been extracted for library construction (data not shown). (amino acids D32, S33, G34) (Table 1, case no. 5). These They included the genes encoding the chitinase 3-like 1 tumor specimens were suitable tools to study the altered (gene ID 1116), the serine protease inhibitor Kazal- gene expression owing to aberrant Wnt signaling from type 1 SPINK1/TATI (gene ID 6690), ubiquitin D the benign to the malignant stages of the liver tumor. (gene ID 10537) and the secreted phosphoprotein 1 or Using the suppression subtractive hybridization tech- osteopontin (gene ID 6696) (Table 2, Supplementary nique, two libraries were constructed and screened on Information). the basis of the overexpression in the respective tester Two genes were found to be common to both tissue. One was expected to be enriched in adenoma- libraries. They encoded REG3A (cloned under different specific genes because adenoma cDNAs have been names HIP/PAP/REG-III, gene ID 5068) (Iovanna subtracted by sequences expressed in the non-tumoral et al., 1991; Lasserre et al., 1992; Dusetti et al., 1994) tissue; conversely, the other was expected to be more and REG1A (gene ID 5967) (Terazono et al., 1988). The

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 601 Table 1 Summary of b-catenin gene alterations in HCC Case Risk factor Nontumoral liver b-catenin mutation b-catenin immunostaining

1* HBV, HCV Cirrhotic D32N ND 2 0 Noncirrhotic delS45-I80 Cytoplasmic/nuclear 31 0 Noncirrhotic delG34-H36 Cytoplasmic/nuclear 41 0 Noncirrhotic S33C Cytoplasmic/nuclear 5 0 Noncirrhotic delD32-H36 Cytoplasmic/nuclear 6* HCV Cirrhotic T41I ND 7 Hemochromatosis Cirrhotic S45F ND 8* HCV, alcohol Cirrhotic G34E ND 9* 0 Noncirrhotic S33A Cytoplasmic/nuclear 10 HBV, HCV Noncirrhotic S45PCytoplasmic/nuclear 11* HCV Cirrhotic S45PND 12 HCV Cirrhotic S37A Membranous 13 HCV Cirrhotic T41A Cytoplasmic/nuclear 14 HBV Cirrhotic — Membranous 151 0 Noncirrhotic delD6-K133 Membranous 16 Alcohol Cirrhotic G34R Cytoplasmic/nuclear 17* 0 Noncirrhotic — ND 18 Alcohol Cirrhotic — Membranous 19* HBV, HCV Cirrhotic — ND 20* HBV Noncirrhotic D32G ND 21 0 Noncirrhotic S33PCytoplasmic/nuclear 22 HCV Cirrhotic D32Y ND 23* Alcohol Cirrhotic — ND 241 0 Cirrhotic S33F Cytoplasmic/nuclear 25* HBV, alcohol Cirrhotic — ND 26 HBV Noncirrhotic — Membranous 27 HBV Cirrhotic — Membranous 28 HCV Noncirrhotic delA39-S47 Membranous 29* HBV, HCV Cirrhotic — ND 30 HBV Cirrhotic — Membranous 31* HBV Cirrhotic — ND 32* 0 Noncirrhotic — Membranous 33 Alcohol Cirrhotic — Membranous 34* HBV Noncirrhotic — ND 35* 0 Noncirrhotic — ND 36* HCV Cirrhotic — ND 37 Alcohol Noncirrhotic T41I Membranous 38 HCV Cirrhotic — Membranous 39* Alcohol Cirrhotic — ND 40 Alcohol Noncirrhotic — Membranous 41 HCV Cirrhotic — Membranous 42 0 Noncirrhotic — Membranous

Asterisks indicate the patients previously described in Christa et al. (1999) and circles those reported in Terris et al. (1999). ND ¼ not done

upregulation of REG3A and REG1A gene expression involvement in liver tumorigenesis. The demonstra- was confirmed in the tissues from which RNA had been tion that they were strongly induced in a b-catenin- extracted for library construction by Northern blot activated liver adenoma and carcinoma suggests a analysis and immunohistochemistry. REG3A and common regulation of these genes by the b-catenin REG1A were not expressed in the non-tumoral liver pathway. and were strongly induced both in adenoma and HCC (Figure 2a). A strong immunoreactivity was observed REG3A mRNA is induced by activation of b-catenin for REG3A protein both in the adenoma and HCC signaling in the Huh7 hepatoma cell line (Figure 2b); it was less pronounced for REG1A for We used the Huh7 hepatoma cell line, which has a wild- which rare foci of stained hepatocytes were detected in type b-catenin status, to test whether the REG3A gene both components (Figure 2c). responds to Wnt signaling in vitro. Wnt signaling was Interestingly, REG3A and REG1A have similar activated in Huh7 cells using lithium chloride (LiCl), organization and are tandomly clustered on chromo- which inhibits GSK3-b and mimics Wnt signaling some 2p12; they encode proteins that are members by stabilizing b-catenin in its unphosphorylated form of the REG family with substantial (Stambolic et al., 1996). Stimulation of Wnt signaling by (Zhang et al., 2003). Their products are secretory LiCl was monitored by transfecting a TCF reporter proteins of the C-type superfamily that are (pTOPFLASH) and a control plasmid containing involved in liver and pancreatic regeneration scrambled TCF-binding sites (pFOPFLASH). Cotrans- and proliferation. Consequently, they are candidate for fection of Huh7 with pTOPFLASH and an oncogenic

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 602 inhibitory effect on REG3A expression at 72 h (Figure 3b). These results are consistent with the REG3A gene being a downstream target of the Wnt/ b-catenin pathway.

Analysis of REG3A and REG1A expression in human HCCs We tested the relationship between the overexpression of REG3A and REG1A and the activation of the Wnt pathway in vivo, by analysing their expression in a panel of HCCs with known b-catenin status (Table 1). These series comprises the HCC tumor from which the REG3A gene was initially cloned (Lasserre et al., 1992); we characterized a posteriori this tumor for its b-catenin status (Table 1, case no. 9) and identified a point mutation altering the S33 phosphorylable residue (S33A) of b-catenin. Accordingly, immunohistochemis- try revealed that this tumor accumulated b-catenin in the nuclei of hepatocytes (Table 1). Of 42 HCCs, 21 (50%) had point mutations or deletions altering target residues for phosphorylation by CK1/GSK3-b, essential for the destruction of b-catenin (Table 1). Of the 21 mutated HCC, 17 were point mutations, the remaining five were deletions that had Figure 2 Expression of the REG3A and REG1A genes in the arisen in a non-cirrhotic liver (Table 1). tumors used for the SSH procedure. (a) REG3A and REG1A mRNAs in RNAs used for the subtracted libraries were confirmed The expression of REG3A and REG1A genes by Northern blotting. NT: non-tumoral; Ad: adenoma; HCC: was analysed by real-time RT–PCR. The overexpression hepatocellular carcinoma. Blots were standardized with a human of REG3A and REG1A was observed in most of the GAPDH probe. (b) Immunostaining for REG3A in the adenoma HCC analysed (Figure 4a). In contrast, REG3A and HCC components (magnification  100). (c) Immunostaining and REG1A were undetectable in non-tumoral livers for REG1A in the adenoma component (magnification  100). and normal livers (data not shown). REG3A and REG1A gene expression correlated with b-catenin mutation status (P ¼ 0.024 and 0.035, respectively) b-catenin (DN89b-catenin) led to a 10-fold increase of (Figure 4a). The 13 samples expressing REG3A the basal reporter activity, whereas the increase was most strongly carried b-catenin mutations. The seven approximately 80-fold when cells were treated for 48 h tumors expressing REG1A most strongly had b-catenin with 20 mM LiCl; no activation of pFOPFLASH activity mutations. Interestingly, six of them displayed a strong was observed with either activated b-catenin or with REG3A expression, consistent with coregulation of LiCl (Figure 3a). Activation of the pTOPFLASH these two genes. We tested the expression of these activity following LiCl treatment was substantially genes in benign liver tumors. No expression was reduced both by cotransfection with the plas- detected in focal nodular hyperplasias (n ¼ 8) in which mid DNTCF4 encoding the dominant-negative form no b-catenin mutation was found. Among the 11 of TCF and transfection of siRNA directed against adenomas analysed, expression of these genes is only b-catenin (Verma et al., 2003) (Figure 3a); scrambled detected in the three cases exhibiting b-catenin gene siRNA had no effect on the pTOPFLASH activity (data deletions (Figure 4b). not shown). A subset of HCCs was then analysed by immunohis- REG3A mRNA was barely detectable by real-time tochemistry. In all, 10 showed a nuclear and/or a RT–PCR in untreated cells. REG3A mRNA was cytoplasmic accumulation of b-catenin, and all 10 slightly induced after 24 h of 20 mM LiCl treatment exhibited b-catenin mutations (Table 1). REG3A and gradually increased until 72 h of LiCl treatment expression differed among the cases analysed: few cases, (Figure 3b). This was confirmed by immunocytochem- all strongly positive for b-catenin, exhibited a marked istry for b-catenin and REG3A: the amount of REG3A and diffuse labeling in more than 50% of tumoral cells; protein had clearly increased after 72 h of treatment most showed weak and heterogeneous staining, for (Figure 3c). REG1A mRNA was undetectable in example, in the HCC with intratumoral heterogeneous untreated Huh7 cells even after LiCl treatment. nuclear staining for b-catenin and REG3A expression To confirm that b-catenin can effectively regulate was strictly confined to these positive areas (Figure 5a). REG3A expression, we transfected siRNA against Owing to the weakness of the REG1A antibody, in situ b-catenin from the beginning of the LiCl treatment. hybridization was performed on some HCC samples and We confirmed that these previously described siRNA on one adenoma sample, which display the strongest (Verma et al., 2003) strongly inhibit b-catenin signaling expression by real-time RT–PCR (Figure 5b). Except a induced by LiCl (Figure 3a) and that they had a clear weak signal around the bile ducts, no expression is

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 603

Figure 3 REG3A mRNA is induced by LiCl treatment in the Huh7 hepatoma cell line. (a) Histogram showing the relative luciferase activity calculated as with a multiple of the activity in Huh7 transfected with pTOPFLASH. Cells were transfected with 0.25 mg of pTOPFLASH (or pFOPFLASH) and 0.5 mg of the DN89b-catenin plasmid or 1.75 mgofDNTCF4 plasmid or 150 nM siRNA (final concentration). Means7s.d. are presented. (b) Histogram showing on a decimal scale the relative gene expression of REG3A measured by real-time RT–PCR in Huh7 cells during LiCl treatment. Relative gene expression is expressed as multiples of the basal value as described in Materials and methods. Huh7 cells were transfected with 150 nM siRNA (final concentration) as indicated. Means7s.d. are presented. (c) Immunocytochemistry for b-catenin and REG3A in Huh7 cells after 24 and 72 h of LiCl treatment. Delocalization of b-catenin was first detectable within 24 h of treatment. observed in the non-tumoral liver for REG3A (data not Analysis of REG3A and REG1A expression in HB and shown) and REG1A (Figure 5b). In situ detection of the colon adenomas REG1A mRNA demonstrating its overexpression was We assessed the responsiveness of REG1A and REG1A further confirmed at the protein level using Western blot to Wnt signaling, by analysing the expression of these analysis (Figure 5c). These results provide strong two genes in HB, a pediatric liver tumor for which a evidence for the regulation of REG3A and REG1A high prevalence of b-catenin mutations has been genes by b-catenin in human liver tumors. reported (Buendia, 2002). Among 28 HB cases analysed,

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 604

Figure 5 Analysis of expression of REG1A and REG3A. (a) Analysis of expression of REG3A by immunohistochemistry. Figure 4 Quantitative analysis of the expression of REG3A and In an HCC exhibiting heterogeneous nuclear staining for b-catenin, REG1A genes in HCCs and in b-catenin mutated adenomas. the expression of REG3A protein was strictly confined to the areas (a) Expression of REG3A and REG1A genes in HCCs by real-time of tumoral hepatocytes displaying cytosolic and nuclear b-catenin RT–PCR. Histograms show on a logarithmic scale the relative gene staining (magnification  100). (b) Detection of REG1A mRNA expression of REG3A (P ¼ 0.024) and REG1A (P ¼ 0.035) by in situ hybridization. Sections were counterstained with Nuclear measured by real-time RT–PCR in a panel of HCC with (black Fast Red. In situ hybridization shows strong (case nos. 3 and 5) and bar) and without (white bar) b-catenin mutations. Relative gene moderate (case no. 2 and adenoma no. 1) expressions of REG1A expression is expressed as multiples of the control value as by tumoral hepatocytes. The non-tumoral liver does not exhibit described in Materials and methods. (b) Expression of REG3A any expression except a faint labeling of bile ductular cells as and REG1A genes in b-catenin-mutated adenomas by real-time indicated by the arrow (adenoma no. 1) (magnification  200). RT–PCR. The b-catenin gene mutations are mentioned. Histo- (c) Analysis of REG1A protein by Western blot. Total cell extracts grams show on a logarithmic scale the relative gene expression of from adjacent non-tumoral and tumoral livers were analysed using REG3A and REG1A measured by real-time RT–PCR in three SDS–PAGE and immunoblotted with anti-REG1A antibody adenomas displaying b-catenin gene deletions. Relative gene (Rechreche et al., 1999). Coomassie Blue staining of proteins is expression is expressed as multiples of the control value as shown. described in Materials and methods.

19 tumors (66.5%) carried a mutated b-catenin gene. As probably reflects context-dependent differences between shown above for HCC, there was a strong correlation tumor cells of different origins (Figure 6a). between REG3A (P ¼ 0.006) and REG1A (P ¼ 0.026) Concomitant upregulation of REG3A and REG1A gene expressions and the presence of b-catenin muta- gene expression during colorectal carcinogenesis has tions (Figure 6a). The level of expression of REG3A in been reported (Rechreche et al., 1999; Macadam et al., HBs was similar to that in HCCs, but mean REG1A 2000). In addition, REG1A mRNA is overabundant in expression was much lower in HBs than in HCCs. This colon adenomas from FAPpatients (Buckhaults et al.,

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 605 Takayasu et al., 2002; Bioulac-Sage et al., 2003) unlike the frequent biallelic inactivation of the gene TCF1 encoding hepatocyte nuclear factor 1 found in 50% of adenomas (Bluteau et al., 2002). In our case, the b-catenin mutation found in the adenoma was an interstitial deletion involving b-catenin exon 3. Interest- ingly, the b-catenin mutations in hepatic adenomas described by Chen et al. (2002) were also deletions. Missense mutations are prevalent in human HCC (our results and for a review (see Buendia, 2000), but in our collection, we found five deletions in HCC occurring in normal, non-cirrhotic liver. HBs are liver tumors occurring in normal liver and show a high prevalence of b-catenin deletions (Takayasu et al., 2001; Buendia, 2002; Taniguchi et al., 2002). The significance of these observations remains unclear, and it may be of interest to compare tumors from apparently morphologically normal liver displaying b-catenin deletions with those bearing b-catenin point mutations and from a cirrhotic context to determine differences concerning pathogeni- city and tumoral development. Using differential screening, we isolated two genes of the growth-promoting lectin (REG) family, the REG3A Figure 6 Analysis of the expression of REG3A and REG1A and REG1A genes. They were overexpressed in primary genes in other tumors displaying activation of the Wnt pathway. liver tumors bearing b-catenin mutations and were (a) Quantitative analysis of REG3A and REG1A gene expression probably target genes of the genetically altered b-catenin in HBs by real-time RT–PCR. Histograms show on a logarithmic signaling in these tumors. scale the relative gene expression of REG3A (P ¼ 0.006) and REG1A (P ¼ 0.026) measured by real-time RT–PCR in a panel of We provide various evidence that REG3A and HBs with (black bar) and without (white bar) b-catenin mutations. REG1A genes are downstream targets of the Wnt Relative gene expression is expressed as multiples of the control pathway. No expression was detectable in normal and values as described in Materials and methods. (b) In a case of non-cancerous liver; expression of these genes was familial adenomatous polyposis, REG3A and REG1A expression significantly associated with HCCs and HBs with a were detected in cells expressing cytosolic and nuclear b-catenin (magnification  200). mutated b-catenin status. Analysis of polyps from FAP patients with constitutively active Wnt pathway showed that both genes were expressed in epithelial cells strongly expressing activated b-catenin. REG3A expression was 2001). These tumors display biallelic inactivation of the induced in Huh7 cells by LiCl treatment, treatment that adenomatous polyposis coli gene (APC), leading to the mimics Wnt signaling, and this expression was abolished stabilization and accumulation of wild type b-catenin. by siRNA directed against b-catenin. However, we We therefore searched for a correlation between the could not demonstrate that b-catenin directly regulates stabilization of b-catenin and the upregulation of REG3A promoter activity: REG3A promoter con- REG3A and REG1A in polyps from FAPpatients by structs containing up to 2000 bp of regulatory sequences immunohistochemistry. Among 12 adenomas from four with several LEF/TCF sites showed no responsiveness patients, eight displayed perfect confinement of the to b-catenin signaling in transfection experiments REG3A and REG1A proteins to adenomatous cells (unpublished data). Albeit indirect, the regulation of where a strong activation of b-catenin was detected REG genes expression by Wnt signaling is reinforced by (Figure 6b). a recent clinical study, which reports in a large series of 265 HCC a strong association between b-catenin mutation and overexpression of REG1A (P ¼ 0.00005) Discussion and REG3A (PAP) (Po0.00001) (Yuan et al., 2005). REG3A and REG1A belong to the C-lectin family To isolate new b-catenin target genes possibly involved and are tandomly clustered on 2p12. They share in b-catenin-induced liver tumorigenesis, we used a rare structural and some functional properties and are case of HCC that arose in a pre-existing adenoma, and involved in injury, inflammation, diabetes and carcino- both developed on non-fibrotic liver without clinical genesis (Zhang et al., 2003). REG3A was first isolated as evidence of glycogenosis. Both components, adenoma a pancreatic secretory protein implicated in pancreatic and HCC, display the same b-catenin mutation that was regeneration, and then identified in human HCC probably the tumorigenic initiating event. (Iovanna et al., 1991; Lasserre et al., 1992; Dusetti Hepatocellular adenoma is a benign liver neoplasm et al., 1994). In the exocrine pancreas, REG3A protein is occurring in morphologically normal liver. Mutations of associated with pancreatic acinar cell protection from the b-catenin gene are rare events (Chen et al., 2002; oxidative stress and TNF-a-induced pancreatic stress

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 606 (Ortiz et al., 1998; Malka et al., 2000). It has also been satisfied the requirements of the local ethics committee. The recently reported to behave as a mitogenic factor samples selected for suppression subtractive hybridization towards hepatocytes, and an antiapoptotic molecule analysis were from a 40-year-old man who had developed a against TNFa-mediated hepatocyte apoptosis (Simon HCC on a pre-existing adenoma. Serologies for HBV and et al., 2003). These effects appear to be mediated in part HCV were negative and the non-tumoral liver did not show any signs of fibrosis or glycogenosis. HCC RNA was extracted through the activity of the protein kinase A (Simon from 25 tumor samples and matched non-tumoral tissue et al., 2003). The REG1A gene encodes a protein where available, obtained at Cochin Hospital (Paris, France); secreted by the exocrine pancreas, which has been 17 HCC RNA samples have been reported previously (Christa associated with islet cell regeneration and diabetogenesis et al., 1999). In all, 28 frozen samples of HB obtained by (Terazono et al., 1988; Watanabe et al., 1994). Thus, the surgical liver resection in different French hospitals were used; REG proteins have been reported to be involved in 23 of these patients had received preoperative chemotherapeu- various biological functions including proliferation, and tic treatments (Perilongo et al., 2004). All samples were resistance to apoptosis during regeneration, inflamma- characterized for the presence of mutations of the b-catenin tion and tumorigenesis, but nevertheless, their exact gene by PCR with genomic DNA or cDNA as described biological function(s) remains to be established. Our previously (de La Coste et al., 1998). observation that both REG3A and REG1A proteins are markedly induced in HCC and HB but are not expressed Cell line, transfection and reporter assays in normal liver suggests that these REG proteins may be The hepatoma cell line Huh7 was maintained in DMEM with involved in human hepatocarcinogenesis. Indeed, it has 10% (vol/vol) calf serum. Transient transfections were been reported that the overexpression of the mouse performed when cells were 60–70% confluent in 12-well plates using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, ortholog of REG1A under the rat glucagon promoter USA). The total amount of transfected DNA (2 mg/well) was led to the development of rare tumors, including HCCs, kept constant by adding pcDNA3. A TK-Renilla plasmid in transgenic mice consistent with tumor-promoting (10 ng) was included in each transfection for monitoring the activity (Yamaoka et al., 2000). Furthermore, REG1A transfection efficiency. Just before transfection, the culture and REG3A have recently been found overexpressed medium was changed for media supplemented with LiCl during rodent liver regeneration after hepatectomy 20 mM or sodium chloride (NaCl) 20 mM. Cells were lysed 48 h (White et al., 2005). The REG proteins are associated after transfection and the luciferase and Renilla activities were with inflammation and regenerative processes and assayed using Promega’s Dual Luciferase Reporter Assay therefore we searched for an association between REG (Promega, Madison, DI, USA). All experiments were per- expression and clinicopathological features, chronic formed in duplicate and repeated at least three times. The DN89b-catenin-pCAN expression vector was kindly provided viral infection or alcohol abuse, but no statistically by P Polakis (San Francisco, USA). pTOPFLASH, pFOP- significant correlation was found. FLASH and the expression plasmid encoding the dominant- The expression of REG1A has been also observed in negative form of TCF (pDNTCF4) were kindly provided by human cholangiocarcinoma and in gastric and colorectal H Clevers (Utrecht, The Netherlands). Double-stranded cancers (Buckhaults et al., 2001; Harada et al., 2001; siRNAs were purchased from Proligo (Paris, France). For b- Fukui et al., 2004). In gastric and colorectal tumors, the catenin, we used the siRNA described previously (Verma et al., expression of this gene is closely correlated to carcinoma 2003); the control siRNA was 50-GGCAUAGAUGUAGCU- invasiveness (Macadam et al., 2000; Dhar et al., 2004). GUAAdTdT-30 and its complement, leaving a dTdT 30 Whether expression of REG1A and REG3A during overhang on both strands. siRNA was transfected into cells hepatocarcinogenesis is a marker of poor prognosis is according to the manufacturer’s instructions with Oligofecta- mine (Invitrogen, Carlsbad, CA, USA) to result in a final still unknown. However, a recent clinical data report RNA concentration of 150 nM, and cells were harvested after 2 their differential expression in HCC according to the and 3 days, respectively. To determine the effects of siRNA on tumor progression: REG3A would be correlated to a reporter constructs, the cells were transfected 6 h after siRNA favorable prognosis, whereas expression of REG1A, transfection with either pTOPFLASH or pFOPFLASH as alone or in combination with REG3A, is associated with described above, and the cells were harvested after 48 h to a poor outcome (Yuan et al., 2005). This observation is assay luciferase activity. close to that found in colorectal carcinogenesis where coexpression of these REG genes appears closely linked Suppression subtractive hybridization (SSH) to the adverse outcome of the disease (Macadam et al., For subtractive hybridization, poly(A) þ mRNA was purified 2000). This overexpression may also account for the using the Micro-FastTrack mRNA isolation kit (Invitrogen, neoplastic progression in HCC with mutational defects Carlsbad, CA, USA) from total RNA of non-tumoral, in b-catenin. Further studies are required to validate the adenoma and carcinoma components of liver from the patient relationship between the seric production of these described above. Poly(A) þ mRNA (2 mg) were used to proteins and their prognostic value in human HCC. prepare cDNA using the SMART cDNA synthesis amplifica- tion kit (BD Biosciences Clontech, Palo Alto, CA, USA). Subtractive hybridizations were performed using the PCR- Select Subtractive Hybridization kit (BD Biosciences Clontech, Materials and methods Palo Alto, CA, USA) according to the manufacturer’s protocol: cDNA from non-tumoral (driver) and adenoma Tumor samples (tester) tissues and cDNA from adenoma (driver) and All tumor samples were obtained from surgical liver resections. carcinoma (tester) tissues were used. The two subtracted Access to this material was in agreement with French laws and cDNA libraries were subcloned into pGEM-T Easy vector

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 607 (Promega, Madison, DI, USA) and used to transform monoclonal anti-b-catenin (1/100, BD Biosciences) (Cadoret Electromax DH10B cells (Invitrogen, Carlsbad, CA, USA). et al., 2002), monoclonal antiglutamine-synthetase (1/200, BD Approximately 40 000 clones were plated for both libraries and Biosciences) (Cadoret et al., 2002), polyclonal pre-HIP(1/400) were screened with the corresponding subtracted tester and (Christa et al., 1999) and monoclonal anti-REG1A (1/200, a driver probes. After secondary screening, 115 clones from the generous gift from C Figarella, Marseille, France) (Bernard- HCC library and 78 clones from the adenoma library were Perrone et al., 1999). Bound antibody was visualized using the chosen for sequencing. immunoperoxidase protocol (Vectastain ABC kit, Vector Burlingame, CA, USA). RNA extraction, Northern blot and RT–PCR analyses Total RNA was extracted from frozen liver or Huh7 cells using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA). North- Western blotting ern blotting was performed as reported previously (Ovejero Western blots were performed as previously described after et al., 2004). The human REG3A probe has been described separation of total proteins (10 mg) through 14% sodium elsewhere (Christa et al., 1999). The human REG1A probe was dodecyl sulfate–polyacrylamide gel (Ovejero et al., 2004), and produced by RT–PCR with RNA from tumoral colon using probed with anti-REG1A-antibody (1/500, a generous gift specific primers and was further subcloned into pGEM-T Easy from JL Iovanna, Marseille, France) (Rechreche et al., 1999). vector (Promega, Madison, DI, USA) and sequenced. Reverse transcriptions and real-time RT–PCR analysis were performed In situ hybridization as described previously (Cadoret et al., 2002; Ovejero et al., The REG1A sense and antisense probes were obtained as 2004). The relative target gene expression was normalized on described above after subcloning into pGEM-T Easy vector the basis of its 18S ribosomal content and to a calibrator that (Promega, Madison, DI, USA). In situ hybridization was was a mix of normal tissue included in each run. The calibrator performed as described previously (Andreu et al., 2005) using a consisted of RNAs extracted from a pool of three colon digoxigenin-labeled probe and detected using an anti-digox- samples. For each sample, the relative gene expression igenin antibody (1/4000) (Roche, Basel, Switzerland). Sections was expressed as x-fold the relative expression of norma- were counterstained with Nuclear Fast Red (Vector labora- lized control tissue. The primer sequences were as follows: REG3A: tories, Burlingame, CA, USA). REG3AF: 50-GGTGAAGAACCCCAGAGGGA-30;REG3AR: 50-CTAGTCAGTGAACTTGCAGA-30. REG1A: REG1AF: 50- ACAGAGTTGCCCCAGGCCCGG-30;REG1AR:50-AGAACT Acknowledgements TGTCTTCACAAGGCA-30.18S:18SF:50-GTAACCCGTT GAACCCCATT-30; 18SR: 50-CCATCCAATCGGTAGTAGCG. We are grateful to B Romagnolo for her help for the SSH Experiments were performed in duplicate for each data point. Mean procedure. We thank PAndreu for in situ hybridization, and F values7s.d. of gene expression from patients according to their Letourneur and S Bulenger for sequencing and all the b-catenin status were compared. The correlation between gene members of our team for helpful discussions. We would like expression and b-catenin status was assessed using Student’s t-test to thank C Figarella and JL Iovanna for providing the (Po0.05). REG1A antibodies. This work was supported by INSERM, the Comite´ de Paris de la Ligue Nationale contre le Cancer, the Immunohistochemistry Association pour la Recherche contre le Cancer, the Ministe`re Immunostaining was performed on 5 mm sections of formalin- De´le´gue´ a` la Recherche (ACI Biologie du De´veloppement et fixed, paraffin-embedded liver. Sections were incubated with Physiologie Inte´grative), the GIS-Institut des Maladies Rares specific antibodies for 1 h. These primary antibodies were and the Ligue Nationale contre le Cancer.

References

Andreu P, Colnot S, Godard C, Gad S, Chafey P, de La Coste A, Romagnolo B, Billuart P, Renard CA, Buendia Niwa-Kawakita M et al. (2005). Development 132: MA, Soubrane O et al. (1998). Proc Natl Acad Sci USA 95: 1443–1451. 8847–8851. Bernard-Perrone FR, Renaud WP, Guy-Crotte OM, Bernard Dhar DK, Udagawa J, Ishihara S, Otani H, Kinoshita Y, P, Figarella CG, Okamoto H et al. (1999). J Histochem Takasawa S et al. (2004). Cancer 100: 1130–1136. Cytochem 47: 863–870. Dusetti NJ, Frigerio JM, Fox MF, Swallow DM, Dagorn JC, Bioulac-Sage P, Laurent-Puig P, Balabaud C, Zucman-Rossi J. Iovanna JL. (1994). Genomics 19: 108–114. (2003). Hepatology 37: 480 (author reply 480-1). El-Serag HB. (2002). Hepatology 36: S74–S83. Bluteau O, Jeannot E, Bioulac-Sage P, Marques JM, Blanc JF, Fukui H, Fujii S, Takeda J, Kayahara T, Sekikawa A, Bui H et al. (2002). Nat Genet 32: 312–315. Nanakin A et al. (2004). Digestion 69: 177–184. Bruix J, Boix L, Sala M, Llovet JM. (2004). Cancer Cell 5: Giles RH, van Es JH, Clevers H. (2003). Biochim Biophys Acta 215–219. 1653: 1–24. Buckhaults P, Rago C, St Croix B, Romans KE, Saha S, Harada K, Zen Y, Kanemori Y, Chen TC, Chen MF, Yeh TS Zhang L et al. (2001). Cancer Res 61: 6996–7001. et al. (2001). Hepatology 33: 1036–1042. Buendia MA. (2000). Semin Cancer Biol 10: 185–200. Iovanna J, Orelle B, Keim V, Dagorn JC. (1991). J Biol Chem Buendia MA. (2002). Med Pediatr Oncol 39: 530–535. 266: 24664–24669. Cadoret A, Ovejero C, Terris B, Souil E, Levy L, Lamers WH Lasserre C, Christa L, Simon MT, Vernier P, Brechot C. et al. (2002). Oncogene 21: 8293–8301. (1992). Cancer Res 52: 5089–5095. Chen YW, Jeng YM, Yeh SH, Chen PJ. (2002). Hepatology 36: Laurent-Puig P, Legoix P, Bluteau O, Belghiti J, Franco D, 927–935. Binot F et al. (2001). Gastroenterology 120: 1763–1773. Christa L, Simon MT, Brezault-Bonnet C, Bonte E, Carnot F, Levy L, Neuveut C, Renard CA, Charneau P, Branchereau S, Zylberberg H et al. (1999). Am J Pathol 155: 1525–1533. Gauthier F et al. (2002). J Biol Chem 277: 42386–42393.

Oncogene Overexpression of REG genes in human primary liver tumors C Cavard et al 608 Macadam RC, Sarela AI, Farmery SM, Robinson PA, Terazono K, Yamamoto H, Takasawa S, Shiga K, Markham AF, Guillou PJ. (2000). Br J Cancer 83: 188–195. Yonemura Y, Tochino Y et al. (1988). J Biol Chem 263: Malka D, Vasseur S, Bodeker H, Ortiz EM, Dusetti NJ, 2111–2114. Verrando P et al. (2000). Gastroenterology 119: 816–828. Terris B, Pineau P, Bregeaud L, Valla D, Belghiti J, Tiollais P Ortiz EM, Dusetti NJ, Vasseur S, Malka D, Bodeker H, et al. (1999). Oncogene 18: 6583–6588. Dagorn JC et al. (1998). Gastroenterology 114: 808–816. Thorgeirsson SS, Grisham JW. (2002). Nat Genet 31: Ovejero C, Cavard C, Perianin A, Hakvoort T, Vermeulen J, 339–346. Godard C et al. (2004). Hepatology 40: 167–176. Verma UN, Surabhi RM, Schmaltieg A, Becerra C, Gaynor Parkin DM, Bray FI, Devesa SS. (2001). Eur J Cancer RB. (2003). Clin Cancer Res 9: 1291–1300. 37(Suppl 8): S4–S66. Watanabe T, Yonemura H, Yonekura H, Suzuki Y, Miyashita Perilongo G, Shafford E, Maibach R, Aronson D, Brugieres L, H, Sugiyama K et al. (1994). Proc Natl Acad Sci USA 91: Brock P et al. (2004). Eur J Cancer 40: 411–421. 3589–3592. Polakis P. (2000). Genes Dev 14: 1837–1851. White P, Brestelli JE, Kaestner KH, Greenbaum LE. (2005). J Rechreche H, Montalto G, Mallo GV, Vasseur S, Marasa L, Biol Chem 280: 3715–3722. Soubeyran P et al. (1999). Int J Cancer 81: 688–694. Yamamoto Y, Sakamoto M, Fujii G, Tsuiji H, Kenetaka K, Satoh S, Daigo Y, Furukawa Y, Kato T, Miwa N, Nishiwaki Asaka M et al. (2003). Hepatology 37: 528–533. T et al. (2000). Nat Genet 24: 245–250. Yamaoka T, Yoshino K, Yamada T, Idehara C, Hoque MO, Simon MT, Pauloin A, Normand G, Lieu HT, Mouly H, Moritani M et al. (2000). Biochem Biophys Res Commun 278: Pivert G et al. (2003). FASEB J 17: 1441–1450. 368–376. Stambolic V, Ruel L, Woodgett JR. (1996). Curr Biol 6: Yuan RH, Jeng YM, Chen HL, Hsieh FJ, Yang CY, Lee PH 1664–1668. et al. (2005). Clin Cancer Res 11: 2568–2575. Takayasu H, Horie H, Hiyama E, Matsunaga T, Hayashi Y, Zhang YW, Ding LS, Lai MD. (2003). World J Gastroenterol Watanabe Y et al. (2001). Clin Cancer Res 7: 901–908. 9: 2635–2641. Takayasu H, Motoi T, Kanamori Y, Kitano Y, Nakanishi H Zucman-Rossi J, Godard C, Boyault S, Tissier F, Grimber G, et al. (2002). Hum Pathol 33: 852–855. Balabaud C et al. (to be submitted). Comparative evaluation Taniguchi K, Roberts LR, Aderca IN, Dong X, Qian C, of Wnt/b-catenin target genes in mutated b-catenin and Murphy LM et al. (2002). Oncogene 21: 4863–4871. Axin1 hepatocellular carcinomas.

Supplementary Information accompanies the paper on Oncogene website (http://www.nature.com/onc).

Oncogene