Clinical and morphological characterizationBetween January 2015 and December 2016, this study evaluated nine cases of HCC, including tumor (T), lateral margin (TM) at the interface with the nonneoplastic liver, and liver distant from the neoplasm (NNL) samples. Six of the cases involved hepatic explants, and three involved hepatic segmentectomy for tumor resection.The cases were selected so that we could study the morphological, molecular and chemosensitivity aspects in pairs. This study was carried out with patients from Portugues Hospital (Salvador, Bahia, Brazil), and the Research Ethics Committee of the Oswaldo Cruz Foundation (Salvador, Bahia, Brazil) approved the protocol (CAAE: 18417813.5.0000.0040). This study was conducted according to the Declaration of Helsinki, and all patients provided informed written consent for the purpose of the research.Under sterile conditions, the samples were immediately transferred from the operating room to the laboratory for sectioning by an experienced pathologist (L.A.R.F.). Three fragments were harvested from each tumor: one was stored in RNAlater (Invitrogen Corporation, USA) for 24 h and frozen in freezer (− 80 °C), the other was fixed in 10% neutral buffered formalin for histological processing, and the other was used to obtain primary HCC cells. The TM and NNLT fragments were also preserved in RNAlater (Invitrogen Corporation, USA) for further analysis.After fixation in 10% neutral buffered formalin, the tumor and nonneoplastic liver samples were dehydrated in alcohol, embedded in paraffin, after xylene baths, and then sectioned to a thickness of 3–5 microns. These sections were stained with hematoxylin and eosin for histological evaluation following WHO recommendations for the classification of histological types and subtypes, as well as histological grading53.Immunohistochemical evaluation of the tumors was performed using the following conventional markers of hepatocellular differentiation: Hep Par-1, arginase, pCEA, and glutamine synthetase. In addition to these markers, CK7, CK19, EpCAM, and CD56 were used to evaluate the presence of hepatic precursor cells in the neoplasms. The expression of GLI1 was evaluated in three patients using a polyclonal antibody. Supplementary Table 2 shows the antibodies used and their suppliers.The histological sections were deparaffinized in xylene and then rehydrated with alcohol. To expose the antigenic epitopes, the sections were subjected to antigen retrieval under moist heat for 45 min. Subsequently, endogenous peroxidase blocking (Peroxidase Blocking Solution™, Dako, Carpinteria, USA) and tissue protein blocking (Protein Blocking Solution™, Dako) were carried out. The slides were incubated with primary antibodies overnight at 4 °C, followed by the application of HRP Link and HRP Enzyme reagents (Advance™, Dako). The reactions were visualized using 3,3′-diaminobenzidine (Dako) and counterstained with Harris hematoxylin. For negative controls, in each reaction, the specific primary antibodies were replaced with an unrelated IgG of the same isotype.In silico studyRNA-seq data for human HCC samples were obtained from the TCGA database (portal.gdc.cancer.gov/projects/TCGA-LIHC). The gene expression data and corresponding clinical information were downloaded from the TCGAbiolinks package (0.18129/B9.bioc. TCGAbiolinks), version 2.1.0 (371 HCC patients and 50 healthy tissues). The edgeR package (version 3.10.5) was used to filter genes with low counts54. All genes with a CPM (counts per million) > one were included for further analysis. Subsequently, a total of 27,644 genes were processed using variance analysis, and the top 25% of the most variant genes (6,911) were selected for coexpression network construction.The WGCNA (version 1.70-3)55 package in R was used to construct a gene coexpression network with 6,911 genes. Coexpression analysis was performed for paired genes using a Pearson correlation matrix. Next, the weighted adjacency matrix was constructed using the power function as follows:$${a}_{ij}= {s}_{ij}^{\beta }$$In this formula, \({s}_{ij}\) represents Pearson’s correlation between gene i and gene j. In addition, β is a soft thresholding approach that leads to a weighted gene coexpression network with a scale-free trait. These values should be chosen in a way that emphasizes strong correlations between genes and penalizes weak correlations. Next, the TOM matrix is calculated from the adjacency matrix to delineate the similarity in nodes, correlating the weighted correlation between two nodes and the other nodes. Finally, to identify genes with absolute highs and cluster them into modules for further analysis, average linkage hierarchical clustering was performed based on TOM-based dissimilarity. The minimum number of genes per module was set to 60.To select the key modules for the analysis, the correlations between module eigengenes, which is the first principal component analysis of the gene module, and clinical traits were assessed. Module significance is defined by the average gene significance for all genes in each module. Modules with higher module significance are biologically relevant for a given condition. Gene significance is reached by a linear relationship between gene expression and clinical traits. In addition, the module membership was measured for all genes and represented its connectivity. All analyses were conducted using R (version 3.6.3)56.GO and KEGG pathway enrichment analyses of the WGCNA modules were performed using the DAVID platform (https://david-d.ncifcrf.gov). To select the most significant GO and KEGG terms, “adjusted” P < 0.01 was used.Molecular biologyThe samples were homogenized (L-Beader, Loccus Biotecnology). Total RNA was extracted using microcolumns (RNeasy Plus Mini Kit, QIAGEN, Tokyo, Japan). RNA purity was evaluated by spectrophotometry (NanoDrop™, Thermo Scientific, Wilmington, USA), while quantification was determined by fluorimetry (QuBit™, Life Technologies, USA). Genomic DNA was eliminated using DNAse I enzyme (Invitrogen Corporation, Carlsbad, CA). For reverse transcription, RNA was processed using the SuperScript® VILO™ cDNA Synthesis Kit (Invitrogen Corporation, Carlsbad, CA). The experiments were performed under conditions free of DNAse/RNAse. TaqMan gene inventoried assays for GLI1 (Hs01110755_m1), GLI2 (Hs01119974_m1), GLI3 (Hs00609233_m1) and PTCH1 (Hs00181117_m1) were used for the qPCR study. After evaluating a total of six (18S, ACTB, B2M, GAPDH, HPRT1, and UBC) reference gene candidates, ACTB (Hs01060665_g1) and GAPDH (Hs02758991_g1) were selected for gene expression normalization. A nonneoplastic liver sample was used as a normal control for calibration purposes.All reactions were developed using an ABI ViiA7 Fast Real-Time PCR System (Applied Biosystems™, Foster City, CA) in a 96-well plate at a total volume of 20 μL, with 8 μL of sample cDNA (20 ng/μL), 1 μL of assay (Applied Biosystems™, Foster City, CA), 10 μL of TaqMan PCR Master Mix (Applied Biosystems™, Foster City, CA) and 1 μL of RNAse-free water. The amplification protocol consisted of an initial cycle at 50 °C (2 min) and 95 °C (10 min) followed by 40 cycles (95 °C) for 15 s and 60 °C (1 min).After amplification and dissociation, relative quantification (RQ) values were obtained using Gene Expression Suite™ v.1.0.3 software (Applied Biosystems™, Foster City, CA) following a comparative method for Cq (2−ΔΔCQ)57.In vitro chemosensitivity assayOne fragment from each tumor was preserved in RPMI-1640 medium (Gibco-BRL, Gaithersburg, MD, USA) supplemented with 10% fetal bovine serum (Life, Carlsbad, CA, USA), 2 mM L-glutamine (Vetec Química Fina, Duque de Caxias, RJ, Brazil) and 50 μg/mL gentamycin (Life, Carlsbad, CA, USA). Tumor fragments were cut with scissors, washed with medium and treated enzymatically for 30 min at 37 °C with 0.5 mg/mL protease, 0.2 mg/mL collagenase type I and 0.2 mg/mL DNase (all from Sigma‒Aldrich Co., Saint Louis, MO, USA). A nylon membrane (100 μm) (Falcon 2350, Becton Dickinson, NJ, USA) was used for filtration. The tumor cell pellet was resuspended in the same complete medium, and the viable cell count was determined by the trypan blue dye (Gibco-BRL, Gaithersburg, MD, USA) exclusion method.To compare the results for primary cells, the well-characterized human liver cancer cell line HepG2 was obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in flasks. HepG2 cells were subcultured every 3–4 days to maintain exponential growth. Treatment with 0.25% trypsin EDTA solution (Gibco-BRL, Gaithersburg, MD, USA) was used to obtain the cell suspension. Additionally, a mycoplasma stain kit (Sigma‒Aldrich Co., Saint Louis, MO, USA) was used to confirm that the cells were free from contamination.Chemosensitivity was evaluated in vitro using the Alamar blue method58. The cells were seeded in 96-well plates for all experiments (100 μL of a suspension of cells at 5 × 105 cells/mL for primary culture or 7 × 104 cells/mL for HepG2 cells in each well). After 24 h, the drugs (all at a concentration of 25 μg/mL) were added to each well and incubated for 72 h. All drugs were dissolved in DMSO at a stock concentration of 5 mg/mL and stored at − 20 °C.Four hours before the end of incubation, 20 μL of a stock solution (0.312 mg/mL) of alamar blue (resazurin, Sigma‒Aldrich) was added to each well. The absorbance at 570 nm and 600 nm was measured using a SpectraMax 190 Microplate Reader (Molecular Devices, Sunnyvale, CA, USA). The inhibition rate (IR) was calculated using the following formula: IR (%) = (1 − T/C) × 100, where T and C represent the absorbance of the drug-treated and control wells, respectively. A mean control tumor absorbance < 0.1 was defined as unsuitable for assessment due to an insufficient number of viable cells as a control, and only samples with a mean control tumor absorbance > 0.1 were accepted. An inhibition rate greater than 50% was defined as the sensitivity value for drug evaluation.Statistical analysisThe Kruskal‒Wallis test and Dunn’s posttest were employed to compare three or more groups, and Spearman’s rank correlation coefficient was used to evaluate correlations between two variables. The overall survival of HCC patients was calculated using the Kaplan‒Meier method and the Cox regression model, and the log-rank test was used to compare survival curves. DFS was calculated with the survival analysis tool KMplotter (http://kmplot.com/analysis/). The results were considered statistically significant when P < 0.05. GraphPad Prism 6.03 was used for statistical analysis (GraphPad Software Inc., San Diego, USA).Ethics statementThe Research Ethics Committee of the Oswaldo Cruz Foundation (Salvador, Bahia, Brazil) approved the protocol (CAAE: 18417813.5.0000.0040), which was conducted according to the Declaration of Helsinki. All included patients provided informed written consent for the purpose of the research.
