Page 14 - Read Online

Nishioka et al. Hepatoma Res 2018;4:1 I http://dx.doi.org/10.20517/2394-5079.2017.46 Page 3 of 12

METHODS
Tissue samples
Forty patients diagnosed with BCLC stage A HCC who were referred to a single medical center for primary
treatment of HCC by partial hepatectomy were prospectively recruited to participate in an institutional
review-board approved clinical research study with written informed consent. All patients were deemed
clinical candidates for hepatic resection by an attending surgeon, and a separate informed consent process
for surgery was completed before study recruitment.

Whole transcriptome analysis
At the time of surgery, tumor and adjacent non-tumor samples were taken from the resection specimen
and conserved in separate containers with RNA Later medium (Thermo Fisher, Waltham, MA). RNAs were
subsequently extracted from homogenized frozen liver tissue lysates in RLT Plus buffer with the All Prep
DNA/RNA Mini kit (Qiagen, Valencia, CA). The isolated RNAs were then stored at -80 ˚C until analysis.

The analytical quality of the total RNAs was assayed using a Bioanalyzer with RNA 6000 Nano chips
(Agilent, Santa Clara, CA) prior to use for this study. Isolated RNAs were then processed following the WG-
DASL assay protocol (Illumina Inc., Sunnyvale, California). Resulting PCR products were hybridized onto
the Illumina HumanHT-12 v4 Expression Bead Chips covering over 24,000 transcripts with genome-wide
coverage of well-characterized genes, gene candidates, and splice variants. Arrays were scanned using the
iScanTM instrument and expression levels were quantified using Genome Studio software (Illumina Inc.,
Sunnyvale, CA). The resulting expression data matrix contained 40 columns representing individual tumor
samples and 20,818 rows corresponding to gene expression data.

This gene expression dataset was pre-processed by generalized log2 transformation with background
subtraction, quantile normalization, and row centering. Each sample was annotated with corresponding
clinical data such as age, gender, FIB-4 score, AFP level, and HCC risk factor data, as obtained from clinical
records. All tumor samples, gene expression data, and clinical parameters were de-identified and assigned a
serial number to maintain patient confidentiality.

Tumor classification based on gene expression signature
Tumor molecular classification was based on the Hoshida system, using sub-classification signatures
previously subjected to meta-analysis in 6 different patient cohorts collected from 3 continents (Asia, Europe,
[10]
and North America) . Based on this classification system, samples were categorized by nearest template
[23]
into 3 distinct HCC sub-classes (labeled S1, S2, and S3) . A false discovery rate (FDR) < 0.05 was used as
the statistical criterion for confident sub-class label assignments.

Clinical classification
The histologic diagnosis of HCC was established for each patient by clinical pathology. These diagnoses
were further confirmed in all tumor samples by a single board-certified hepatobiliary pathologist. Tumor
samples were then sub-categorized based on several clinical parameters to be later used as classes for gene
set enrichment analysis (GSEA). These categorizations were based on the distribution of each clinical
parameter for all tumor samples. The clinical parameters to be used as class phenotype labels were selected a
priori. They were age, gender, FIB-4 score, AFP level, and presence of HBV infection. Except for gender and
HBV infection, which are binary, parameters were dichotomized for GSEA based on analysis of dispersion.
AFP levels displayed a bimodal distribution so that a cut-off point between “high” and “low” AFP values
could be made at the histogram trough corresponding to 400 ng/mL. Coincidentally, an AFP cut-off point
[24]
of 400 ng/mL is frequently used clinically as a highly specific cut-off for confirming HCC diagnosis ,
and also frequently serves as a cut-off point for determining eligibility in clinical trials involving agents
with potential selectivity for AFP-producing tumors (e.g. NCT02435433). In contrast, the distribution