Background
Hepatocellular carcinoma (HCC) is the most common primary malignancy of the liver and the second leading cause of cancer death in men worldwide [
1]. In patients with HCC, the prediction of prognosis is more complex compared with other solid tumors since there is no worldwide consensus on the use of any HCC staging system [
2,
3]. Clinical studies demonstrate that only one-third of the newly diagnosed patients are presently eligible for curative treatments [
4] and the 5-year survival after resection for early-stage HCC ranges from 17 to 53 % with recurrence rate as high as 70 % [
5]. Therefore, prognosis estimation and indicators for successful treatment options are critical steps in the management of patients with HCC.
Genes that are commonly dysregulated in cancer are clinically attractive as candidate prognostic markers and therapeutic targets. Previous bioinformatics analyses of gene expression profiles have revealed targets for predicting prognosis and survival in patients with HCC are involved in angiogenesis, cell cycle regulation, invasion and metastasis [
6‐
11]. Although high-throughput genomic technologies have facilitated the identification of cancer biomarkers and improved our understanding of the molecular basis of tumor progression, the most common drawbacks of these studies are a lack of agreement due to the differences across experimental platforms, sample size and quality, inconsistent annotation, ongoing discovery as well as the methods used for data processing and analysis. Moreover, the number of prognostically-informative genes in HCC varies from 3 to 628, with low predictive accuracy, which leads to inherent difficulties in drawing definitive conclusions [
12‐
15]. Therefore, identification of robust biomarker candidates for HCC provides a novel potential link between clinical prognosis and cancer survival rates.
In this study, a meta-analysis was used to obtain a consistent gene expression signature for HCC using the integrating microarray data. The dysregulated genes with potentially high clinical significance were validated by qRT-PCR, among which KLHL21 was the most promising. Suppressing its expression inhibited cell proliferation, migration and invasion in HCC cells. Our analyses identified a novel set of HCC biomarkers with high accuracy, using a combination of molecular techniques and clinical information from patients with HCC. This may lead to potential prognostic and therapeutic applications in the future.
Methods
Data acquisition, inclusion criteria and study strategy
We searched the published microarray datasets from Gene Expression Omnibus (GEO,
http://www.ncbi.nlm.nih.gov/geo/) [
16] and ArrayExpress (
http://www.ebi.ac.uk/arrayexpress/) [
17] up to June 2015, with keyword “hepatocellular carcinoma OR HCC” filtered by organism “Homo sapiens”. To identify new prognostic biomarkers in HCC, the selected microarray datasets must meet the following criteria: (i) both tumor tissues and their adjacent tissues (or normal tissues) were included; (ii) contained contain a large number of patient samples (>50) and high gene coverage (>10,000 filtered genes). After background correction and normalization of raw data, multiple probe sets were reduced to one per-gene symbol using the most variable probe measured by interquartile range (IQR) values across arrays. Significance analysis of microarray (SAM) [
18] was used to determine the differentially expressed genes (DEGs), with a false discovery rate (FDR) <0.001 and 1,000 times permutations.
Functional analysis of DEGs
To investigate the cellular component (CC), molecular function (MF) and biological process (BP) of DEGs, Gene Oncology (GO) enrichment analyses were performed by Database for Annotation, Visualization and Integrated Discovery (DAVID) [
19,
20] and WEB-based GEne SeT AnaLysis Toolkit (WebGestalt). To investigate regulatory network, pathway enrichment analyses were performed by BRB-ArrayTools based on KEGG (
http://www.genome.jp/kegg/) and BioCarta (
http://www.biocarta.com/). In this study, the LS/KS permutation test was used for pathway enrichment and gene-sets with
p < 0.00001 were considered significant. Co-expression analysis of the DEGs was performed with a Spearman correlation coefficient absolute value > 0.75 (
p < 10e-10) by Cytoscape [
21].
Survival analysis
To analyze the correlation between gene expression and clinical relevance, the association between the gene expression levels and survival of patients with HCC was analyzed using the GSE10186 entry. In univariate survival analyses, Cox proportional hazard regression model (Wald test) were used to identify factors important for survival followed by 1,000 times permutation test. In univariate survival analyses, Kaplan-Meier method and the log-rank test were used to compare overall survival curves between high and low gene expression groups. For all statistical analyses, p < 0.05 were considered significant.
Literature confirmation
The DEGs identified from meta-analysis were validated by publications and scientific literature available on PubMed (
http://www.ncbi.nlm.nih.gov/pubmed/?term=). Keyword used, take gene “MYCN” for example, was “(((((survival[Title/Abstract]) OR prognosis[Title/Abstract]) OR biomarker[Title/Abstract]) AND tumor[Title/Abstract]) OR cancer[Title/Abstract]) AND MYCN[Title/Abstract]”.
Cell culture and primary tissues
MHCC97H and HCC-LM3 cells were purchased from the Cell Bank of Type Culture Collection of Chinese Academy of Sciences (Shanghai, China) and maintained according to the supplier’s instructions. Twenty-eight primary HCC tissue samples with paired adjacent normal liver tissue samples were collected and all experimental procedures were approved by the IRB of Third Affiliated Hospital of Third Military Medical University (Chongqing, China). None of the patients had received chemotherapy or radiotherapy before or after surgery. Written informed consent was obtained from all patients or their guardians and all samples were histologically confirmed before analysis.
QRT-PCR analysis
To prepare cDNA, 1 μg total RNA was extracted from cell lines and tissue samples using QIAGEN OneStep RT-PCR Kit. Amplifications of cDNA stocks were performed by qRT-PCR in triplicate using GoTaq qPCR Master Mix (Promega) as described previously [
22‐
24]. In this study, unique primer pairs (Additional file
1: Table S1) used to amplify the selected genes were designed using Primer-Blast at NCBI (
http://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi) and assessed for secondary structure using M-Fold (
http://mfold.rna.albany.edu/). Where possible, the primers were designed to span or include an intron to avoid amplification of genomic DNA and to have similar melting temperatures in the range 56–62 °C. Relative gene expression levels were analyzed by the ΔΔCT method and normalized against
β-actin.
Gene silencing by RNA interference
HCC cells were transiently transfected with small interfering RNA (siRNA) using DharmaFECT (Dharmacon, Lafayette, CO). Twenty-one base pair siRNA duplexes targeting KLHL21 gene (siKLHL21-1: 5′-GTACAACTCAAGCGTGAAT-3′; siKLHL21-2: 5′-TGTCATTGCTGTCGGGTTA-3′) and a standard control (Dharmacon siCONTROL nontargeting siRNA) were synthesized by Dharmacon.
Cell proliferation, migration and invasion assays
For cell proliferation assays, HCC cells were seeded into 96-well plate at a density of 1 × 10
3 cells. The cell proliferation rate was analyzed at different time points (1–5 days) with CellTiter 96® AQueous One Solution Cell Proliferation assay (Promega, Madison, WI) according to manufacturer’s instruction. The absorbance at 490 nm was measured with a microplate reader and the average absorbance values from six wells per group were calculated. Quantitative cell migration and invasion assays were performed using 24-well Boyden chambers (Coring, NY, USA) as described previously [
22‐
24]. The numbers of migrated and invaded cells in six randomly selected fields from triplicate chambers were counted in each experiment under a Leica inverted microscope (Deerfield, IL, USA).
Statistical analysis
Differences in quantitative data between two groups were analyzed using 2-sided paired or unpaired Student t-tests. All of the analyses were performed using SPSS software version 18.0 (SPSS, Chicago, IL, USA). P < 0.05 was considered to be statistically significant.
Discussion
Meta-analysis has been widely used as a powerful method in searching DEGs in various types of cancers [
35‐
39]. In this study, we systematically identify a set of molecular prognostic markers for HCC using meta-analysis. To minimize the limitation from a single microarray dataset, we examined the overlap among many studies using different platforms in an unbiased manner. By comparing gene expression data from 1525 paired samples profiled in the GEO datasets, and by combining molecular and clinical data to reduce false-positive errors, we demonstrate a core gene set with prognostic potential.
Cancer biomarkers are the measurable molecular changes to either cancerous or normal tissues of patients [
40‐
42]. A reliable biomarker can be used for cancer diagnosis, risk and prognosis assessments, and more importantly, some of them can be exploited as therapeutic targets. Therefore, better understanding of the biological significance of such markers and validation of their usefulness are pivotal for developing novel targeted therapies. HCC appears to be characterized by increased glycolysis, attenuated mitochondrial oxidation, and increased arachidonic acid synthesis [
43], suggesting abnormal metabolism in HCC development and progression. In this study, GO analysis, KEGG and BioCarta pathway analyses were performed to determine the roles and pathways of DEGs. These analyses implicate that the expression profiling of metabolism genes was significantly changed in HCC. The deregulated energy metabolism of cancer cells modifies the metabolic pathways and influences various biological processes including cell proliferation. Not surprisingly, the dysregulated genes identified in our study were highly associated with cell cycle pathways.
In order to determine the clinical relevance of the DEGs, survival analysis was performed and 79 DEGs were found to be associated with overall survival. Most of these genes (65.82 %) have prognostic features and strong associations with some cancers. For example,
MYCN is well-studied biomarker for neuroblastoma and inactivation of it results in impaired cell growth and enhanced cell death in neuroblastoma [
44‐
46].
RHEB acts as a proto-oncogene in the appropriate genetic milieu and signaling context, and its overexpression cooperates with
PTEN haploinsufficiency to promote prostate tumorigenesis [
47]. The elevated expression levels of these two genes are also found in our study, suggesting that cancers from different tissues may share common features and these genes can be utilized as pan-cancer biomarkers. The expression levels of
GPC3 are down-regulated to facilitate cell migration, invasion and tumorigenicity in ovarian cancer [
48,
49]. However, our study shows that
GPC3 is an up-regulated gene in HCC, which agrees with other studies [
50‐
53]. These observations indicate that the same gene might exhibit opposite effects on different cancer types, and the genes like
GPC3 cannot be used as pan-cancer biomarkers.
The HR derived from the Cox proportional hazards model provides a statistical test of treatment efficacy and an estimate of relative risk of events. Therefore, understanding of HR of queried gene expression would be helpful in anticancer strategies. Two separate analyses were performed for the genes up-regulated in poor prognosis patients (HR > 1 by the Cox regression) and for those down-regulated in poor prognosis patients (HR < 1). From this analysis, we identified 12 DEGs whose expression levels are associated with significantly higher risk of tumor recurrence, and 4 genes have been reported to be related with survival or prognostic features. For instance,
MAP3K7 controls a variety of cell functions including transcription regulation and apoptosis through mediating the signaling transduction induced by TGFβ and bone morphogenetic protein (BMP) in a broad range of cancers [
54‐
56].
KLHL21 interacts with Cullin3 and regulates mitosis in HeLa cells [
57]. Unlike other family members,
KLHL21 regulates of the chromosomal passenger complex translocation at the onset of anaphase and is required for completion of cytokinesis [
57]. It appears that
KLHL21 is the most promising gene among the 6 validated novel candidates. We identified for the first time that reduced expression of
KLHL21 is associated with decreased cell proliferation rate and invasion potential in HCC cells, although further research is required to fully illustrate the regulatory network and downstream targets of
KLHL21 in HCC development and progression.
Despite the significant body of literature describing predictive or prognostic mRNA profiles for cancer, only a small number are used in current oncology practice. Our study reveals novel biomarkers and molecular signatures related to HCC development and progression, making it possible to objectively evaluate the patient’s overall outcome and translate new molecular information into drug therapy.
Acknowledgement
We are grateful to Dr. Catherine Jauregui for the thorough analysis of manuscript.