HGF/c-Met related activation of β-catenin in hepatoblastoma

SIOPEL Liver tumor clinical trials are international, prospective, clinical trials run under the auspices of the SIOP Liver Tumor Strategy Group (SIOPEL). Our cohort comprises patients prospectively enrolled into the SIOPEL 3 clinical trial, a randomised study which opened in March 1998, designed to evaluate the effectiveness of preoperative chemotherapy for standard risk (SR) HB with either cisplatin (CDDP) alone or in combination with doxorubicin (PLADO). A detailed description of the SR patient cohort, its clinical features, staging and outcome has previously been reported [33]. SIOPEL 3 patients with high risk (HR) HB were all treated preoperatively with SUPERPLADO, a three-drug combination of Cisplatin, Doxorubicin and Carboplatin and the results have been reported [34]. All patients were recruited to the SIOPEL 3 clinical trial with appropriate informed consent. This specific study was reviewed and approved by the New Zealand Health Research Council Multi-regional ethics committee (MREC).

In this study we have accessed a representative cohort of 84 HB patients with clinical, histologic and survival data available for most samples. Both diagnostic and post-chemotherapy samples were available for fourteen patients bringing the total number of samples analysed to 98. In the case of diagnostic samples there was generally just a single formalin-fixed paraffin-embedded (FFPE) tumor block available containing the entire biopsy material on which the diagnosis was made. For each post-chemotherapy case, the most representative FFPE block was identified by examination of slides stained with haematoxylin and eosin (H+E). From the H+E slides, representative tumor and adjacent normal tissue areas were selected by a pathologist (C.M.) for subsequent tissue array construction.

A tissue microarray (TMA) was constructed by depositing a 1 mm core of each tumor or normal tissue into a wax recipient block using the Manual Tissue Arrayer I (Beecher Instruments Inc., Sun Prairie, WI, USA). In cases where tumor heterogeneity was evident, different representative areas of the tumor were sampled for TMA construction. The tissue array block was made in duplicate and 4 μm sections of the TMA blocks were cut for subsequent use in immunohistochemical (IHC) analysis. One TMA section was also stained with H+E for evaluation by pathologists (CM +CT).

The sample cohort consists of 98 samples from 84 patients comprising 62 diagnostic tumour biopsies and 36 post-surgical specimens (both diagnostic and surgical specimens available in 14 cases). Histologic information was available for 91 samples. The tumours were examined centrally and classified as either wholly epithelial (n = 33) or mixed epithelial and mesenchymal (n = 54). One tumour was diagnosed as hepatocellular carcinoma (fibrolamellar type) and one as a small cell undifferentiated (SCUD). The epithelial component was further subtyped as pure fetal (n = 43), embryonal (n = 3) or mixed fetal and embryonal (n = 41). Two tumors were subtyped as macrotrabecular type. Focal anaplasia was seen in three tumors and cholangioblastic features in two tumors. Thirteen cases of osteoid formation were noted in the histology reports with additional osteoid formation in a post-chemotherapy sample that lacked osteoid in the diagnostic biopsy. Teratoid features were noted in seven samples.

Clinical information that classified patients into the two well-defined risk groups was available for 71 patients in our cohort. Twenty-seven of these were high-risk and forty-four were standard risk. Of these 71 patients, nine were born with low birth weight. PRETEXT classification revealed that there were two PRETEXT stage 1 patients, twenty-two stage 2, thirty-one stage 3 and sixteen stage 4 patients. Only two patients had serum AFP levels of < 100 at diagnosis, making them high-risk. Eight and seven patients had portal vein and vena cava involvement respectively, and extrahepatic intra-abdominal disease was seen in three patients also making them high-risk cases. Metastatic disease was present at diagnosis in thirteen children. Relapse or progression in five HR cases resulted in the death of four patients. In the standard-risk group there were six relapses leading to a single death from disease.

Briefly, 4 μm TMA slides were deparaffinized with xylene and ethanol. Antigen retrieval was performed by pressure cooking for 2 minutes in citrate buffer pH6.0. Endogenous peroxidases were blocked with 0.3% hydrogen peroxide and non-specific binding was blocked with normal goat serum. Slides were incubated overnight at 4°C with primary antibodies: Y1234/5-c-Met at 1:300 dilution, Y654-β-catenin at 1:25 dilution and β-catenin at 1:200 (All from Abcam, Cambridge, UK). The EnVision HRP/DAB detection system (Dako, Glostrup, Denmark) was used to visualise the results. Slides were lightly counterstained with haematoxylin. All antibodies were optimized for use in IHC using breast tumour control tissue and the appropriate positive and negative controls were used.

Immunostaining for β-catenin was scored as normal membranous, diffuse or focal cytoplasmic and diffuse or focal nuclear staining. Staining for Y654-β-catenin was scored as negative, cytoplasmic and/or nuclear staining. Staining for Y1234/5-c-Met was scored as positive (cytoplasmic) or negative. Each array duplicate was also stained and the results collated. The staining intensity was noted but not factored, as differing age of donor blocks and variation in fixation methods can impact on staining intensity. The IHC results were analysed in conjunction with two pathologists (CM and CT).

Representative areas of tumour were identified on H+E slides by pathologists and a 1 mm tissue core removed from corresponding areas on paraffin blocks. The RNA was extracted using RecoverALL™ Total Nucleic Acid Isolation kit (Ambion, Austin TX, USA) as per manufacturer's instructions. Normal adjacent tissue was also removed and RNA extracted where it was available in 62 cases.

Samples with the following quality parameters were analysed for CTNNB1 gene mutations: Optical density ratio 260/280 of 1.8 - 2.2 and RNA concentration of > 20 ng/ul using a Nanodrop spectrometer (Thermo Scientific, Wilmington, MA, USA). A 150 bp region of the CTNNB1 gene was amplified that includes the β-catenin regulatory region of exon 3 (codons 32-45) using the following primer pair (B-Cat3/B-Cat2): 5' GATTTGATGGAGTTGGACATGG 3' and 5' TCTTCCTCAGGATTGCCTT 3'. Samples were reverse transcribed and amplified using One-Step RT-PCR kit (QIAGEN, Dusseldorf, Germany) on a DNA Engine Thermal Cyclar (BioRad, Hercules, CA, USA). Reverse transcription was at 50°C for 30 minutes followed by first strand synthesis at 95°C for 15 minutes. 35 cycles of 30 seconds each of denaturation at 94°C, annealing at 52°C and extension at 72°C were carried out. Each reaction contained 1 μl RNA template, 2 μl of enzyme mix, 0.6 mMol of forward and reverse primers, 400 μM of each dNTP, 2.5 mM MgCl₂ in a final reaction volume of 50 μl. RT-PCR products were visualised on a 1.5% agarose gel with ethidium bromide. Amplified RT-PCR products were purified using QIAquick PCR purification kit (QIAGEN) as per manufacturer's instructions. Cycle sequencing was carried out on a GeneAmp^® PCR System 9700 thermocycler using ABI Prism Dye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Foster City, CA, USA) using 20 ng RT-PCR product. Sequencing products were run on an ABI 373A sequencer (Applied Biosystems) and all mutations were verified by sequencing the sense and anti-sense strands. Mutation analysis was carried out using Variant™ Reporter Software (Applied Biosystems) and showed good quality traces spanning the region of interest.

Human hepatoblastoma cells, Huh-6 (JCRB, Osaka, Japan) were routinely maintained in minimum essential media (MEM) containing 10% FBS and penicillin/streptomycin. The human hepatocellular carcinoma cell line Huh-7 (JCRB) was cultured in Dulbecco's minimum essential media (D-MEM) with 10% FBS and penicillin/streptomycin. The cells were serum starved for 24 hours prior to treatment with recombinant human HGF (Invitrogen, Carlsbad, CA, USA) to a concentration of 50 ng/ml for 30, 60, 90 and 120 minutes.

Nuclear and cytoplasmic protein fractions were isolated from the cell lines at the timepoints indicated with the CelLytic™ NuCLEAR™ Extraction kit (Sigma^®, Missouri, USA). The lysate protein concentrations were determined by bicinchoninic acid protein assay using BSA as a standard (Pierce, Rockford, IL, USA). Aliquots of the samples were stored at -80°C until use.

Approximately 20 μg of protein sample were run on NuPAGE 4-12% BisTris gels (Invitrogen) with MES-SDS buffer (Invitrogen) using the Xcell SureLock™ Mini-Cell (Invitrogen). The protein marker used was Precision Plus Protein™ Standards (BioRad). The iBlot Gel Transfer Device (Invitrogen) was used for western blotting of proteins. The filters were probed with anti-Y654 β-catenin (Abcam, 1:150) and anti-β-catenin (Abcam, 1:1000). The filters were stripped with a mild stripping buffer containing 1.5% glycine, 0.1% SDS and reprobed after each blot. The immunoblots were incubated for 1 hour with the appropriate secondary antibodies coupled to horseradish peroxidase followed by exposure to ECL plus chemiluminescence reagents (GE Healthcare Biosciences, Piscataway, NJ, USA) and autoradiography. Immunoblotting with anti-TBP for nuclear proteins and anti-β-actin for cytoplasmic extract was used to confirm equal loading.

We examined total β-catenin protein expression on a HB tissue array using IHC. A total of 87% (85/98) of tumours in our clinical cohort showed aberrant expression of β-catenin in the nucleus and cytoplasm (38/98) or in the cytoplasm alone (47/98) (Figure 1a and 1b). Normal membranous staining alone was observed in seven cases and the remaining six tumours were completely negative for total β-catenin staining. Samples of adjacent normal tissue had a normal membranous β-catenin staining pattern in 46/48 cases available for examination (Figure 1c). The remaining two normal samples showed focal cytoplasmic staining. These results are similar to those published previously in HB studies [18, 35, 36]. However the frequency of mutations in the CTNNB1 gene varies widely in studies of HB, from 13% to 70% [19, 37]. To determine whether aberrant β-catenin protein expression is a result of gene mutation, we identified the frequency and type of CTNNB1 mutations in our cohort.

To identify CTNNB1 mutations we extracted total RNA from corresponding tissue cores of hepatoblastoma. A 150 pb region of the β-catenin regulatory region of exon 3 of the CTNNB1 gene (codons 32-45) was amplified successfully by RT-PCR in 92 of the samples. Lack of amplification in 6 samples may be due to deletion of exon 3 of CTNNB1. We attempted to amplify a region spanning exon 2 to exon 4 in these 6 samples but were unsuccessful. Therefore our estimation of samples containing deletions may be inaccurate. We identified 11 different point mutations in 14 of 98 samples (15%) (Table 1). These are all missense mutations affecting phosphorylation sites in the regulatory region of the gene and have been previously reported [17, 38]. The mutations found, resulted in the following changes at the protein level; 32D > N, 32D > Y, 32D > V, 32D > A, 33S > P, 33S > C, 34G > R, 34G > E, 34G > V, 35I > P, 35I > S, 37S > Y. One HB patient (CCRG 64) showed the same sequence variation (missense 32D > V) in both diagnostic and post chemotherapy tumour samples. RNA from adjacent normal tissue was also analysed from 62 cases including nine tumours that harboured mutations. All of these samples displayed wild type CTNNB1 showing that the mutations found were somatic variants (results not shown). The frequency of CTNNB1 mutations (14/98) and possible deletions (6/98) in our cohort was significantly lower than the frequency of aberrant expression of β-catenin protein and statistical analysis shows no correlation between aberrant β-catenin accumulation and gene mutation/deletion. This prompted us to investigate alternative pathways of β-catenin activation in hepatoblastomas in our patient cohort.

To investigate the possibility of Wnt-independent activation of β-catenin, we analysed our tumour cohort for possible HGF/c-Met related tyrosine phosphorylation of β-catenin. We stained the hepatoblastoma tissue array using an antibody recognising tyrosine 654-phosphorylated β-catenin (Y654-β-catenin). This identified positive staining in the cytoplasm of 82/98 (83%) tumours with an additional 27 (28%) showing nuclear accumulation of Y654-β-catenin. In 78 hepatoblastoma with wild type CTNNB1, 26 (33%) showed nuclear expression of Y654-β-catenin, 44 (56%) showed cytoplasmic staining with only 7 (9%) negative for staining. In contrast, IHC analysis of 20 hepatoblastoma with CTNNB1 mutations or possible deletions showed 5 (25%) were completely negative for Y654-β-catenin (Figure 2a), 14 (70%) had cytoplasmic staining alone (Figure 2b), and only one of 20 (5%) had nuclear expression in addition to cytoplasmic staining (Figure 2c).

Statistical analysis shows a significant correlation between nuclear accumulation of tyrosine-phosphorylated β-catenin and HB tumours with wild-type CTNNB1 (P-value = 0.015).

To verify that tyrosine phosphorylation of β-catenin is specifically due to activation of the HGF/c-Met pathway we examined the expression of tyrosine 1234 and 1235-phosphorylated c-Met. These tyrosine residues become auto-phosphorylated specifically in response to HGF ligand binding. Eighty-one tumour samples (82%) were positive for Y1234/5-c-Met staining (Figure 3a) and the remaining 17 samples were negative (Figure 3b). A single tumour sample showed a distinct nuclear staining pattern with the antibody to Y1234/5-c-Met (Figure 3c). Statistical analysis showed a 70% correlation between Y1234/5-c-Met and Y654-β-catenin expression (r = 0.7). No correlations between staining patterns and histologic subtypes were found with any of the antibodies used.

To corroborate our immunohistochemistry findings on tissue array, we analysed in vitro total β-catenin and Y654-β-catenin protein expression in response to exposure to HGF in two liver tumour cell lines, one with and one without mutation in CTNNB1 (Huh-6 and Huh-7 respectively). To determine their CTNNB1 status, the Huh-6 and Huh-7 cell lines were analysed for CTNNB1 mutations in exon 3 using RT-PCR and sequencing as outlined above. The hepatoblastoma cell line, Huh-6, carried a missense mutation of G34G > V, a known variant of CTNNB1 while the hepatocellular carcinoma cell line, Huh-7, was wild type CTNNB1 (Figure 4).

Analysis of immunoblots using the Y654-β-catenin allowed us to determine how much of the observed nuclear β-catenin expression may be due to activation by HGF/c-Met rather than an activating CTNNB1 mutation. No Y654-β-catenin was seen in any untreated cell fraction, in either the wild type or mutant cell lines. However, upon treatment with HGF the wild type Huh-7 cell line showed significantly more β-catenin expression in the nuclei and cytoplasm compared to Huh-6 (Figure 5).