Background
Hepatocellular carcinoma (HCC) is one of the most frequently occurring malignant tumors worldwide [
1]. Risk factors of HCC are well recognized including gender, infection by hepatitis B virus or hepatitis C virus, cirrhosis, metabolism diseases, toxins, excess alcohol consumption, and smoking. HCC varies with wide geography, and is more prevalent in Asia, Africa, and southern Europe. It has been well defined that experiencing surgery for early HCC patients could achieve a higher curative resection rate (80.5%) [
2], and finally have a better survival rate. However, patients with early HCC frequently manifest non-typical symptoms, hence, most of patients are diagnosed with advanced HCC when seeing a doctor, resulting in a low 5-year survival rate, ranging from 50 and 70% [
3]. Therefore, developing biomarkers for early diagnosis is being emphasized to prolong survival in patients with HCC.
Over the last decades, large efforts have been made to promote the early diagnosis of HCC. Alpha-fetoprotein (AFP) has been the most commonly used tumor biomarker in the liver, testicles, and ovaries [
4]. Highly sensitive and specific biomarkers need to be developed in HCC diagnosis. Glypican-3 (GPC3), a membrane-associated heparan sulfate proteoglycan, is up-regulated in HCC. Additionally, GPC3 involved in hippo pathway to exert its function in HCC cell proliferation. GPC may be applied in clinical practice as a novel diagnostic biomarker [
5].
Additionally, some researchers have attempted to employ prognostic markers for predicting HCC recurrence. Villa E et al. detected whole genome microarray expression profiling of 161 HCC samples, and revealed that five-gene signature (
ANGPT2,
NETO2,
NR4A1,
DLL4,
ESM1) was able to predict fast growth and worst survival of HCC patients [
6]. The exploration of prognostic markers may facilitate individualized therapies.
Recently, detection of genome-wide gene transcripts expressed in a given tissue type is becoming more and more feasible with advent of high-throughput technologies, such as microarray and RNA-seq. The application of microarray-based gene expression profiling has produced tremendous information, and provided mechanistic insights into the oncogenic process of HCC [
7]. However, although many microarray studies of HCC have been performed [
8‐
11], each of study holds a somewhat different view due to the heterogeneity caused by the variety in clinical samples, platform, analytical approach, etc. Toward this end, an integrated analysis of seven HCC gene expression datasets was conducted to identify differential expressed genes (DEGs) between tumor and normal tissues, revealing a common biological thread that linked the disparate microarray studies. Ten genes were selected for further real time polymerase chain reaction (RT-PCR) and TCGA database validation, to prove the credibility of this integrated analysis. We expected our study would be of some value for the future diagnosis and therapy of HCC in clinic.
Discussion
It is generally accepted that the altered gene expression pattern of a cancer tissue should be associated with the initiation and maintenance of the malignant phenotype. Previous studies have identified several HCC gene expression profiles [
18‐
21]. However, there wasn’t a common pattern among disparate studies for HCC. While in this study, we integrated different microarray studies to identify a precise gene expression profile for HCC with more statistical power supported by large sample size. In the current study, an integrated analysis of seven HCC microarray datasets was conducted, and showed that 1167 DEGs were identified, among which 628 genes were up-regulated and 539 genes were down-regulated. These genes mainly participated in the process of cell cycle, oocyte meiosis, and oocyte maturation mediated by progesterone.
In the current study, further annotation and PPI network analysis of the 20 most significant DEGs were conducted. Most of the 20 genes were involved in the pathways of cell cycle, cytokines-cell factor receptor interactions, and intracellular signaling cascades, and their involvements in HCC have also been reported [
22‐
26]. The functions of the 20 genes were in accordance with the results of GO and KEGG analysis. Three genes, including
CCT3,
NDC80, and
ASPM were proved to be highly connected in the PPI network.
CCT3, a subunit of CCT cluster, plays a role in assisting the folding of proteins involved in important biological processes.
CCT3 was found to display a significantly different gene expression level in HCC compared to adjacent non-malignant liver tissues, arising from the occurrence of the amplicon 1q21-q22 [
27], which is consistent with our result of RT-PCR validation. In addition, other genes’ expression status detected by RT-PCR was totally in accordance with the result of integrated analysis, suggesting that the bioinformatics method of integrated analysis was credible.
ASPM was highly expressed in fetal tissues but lowly in most adult tissues. Our result and previous evidences [
23] found that
ASPM and
NEK2 mRNA was over-expressed in HCC. Moreover, we found that
ASPM,
NEK and
CCT3 over-expression present significant association with overall survival of HCC patients based on TCGA validation, predicting enhanced invasive/metastatic potential of HCC and higher risk of early tumor recurrence.
ASPM,
NEK and
CCT3 may be applied as potential prognostic biomarkers for HCC.
CAP2 overexpression was also discovered in our study, and
CAP2 has been suggested as a candidate biomarker of HCC owing to elevated level in the serum of HCC patients [
28].
Among the 10 most significantly down-regulated genes,
DCN, an extracellular matrix proteoglycan, has important biological functions in growth, development and diseases. Loss of the decorin gene, which are known to interfere with cellular events of tumorigenesis mainly by blocking various receptor tyrosine kinases such as EGFR, Met, IGF-IR, PDGFR and VEGFR2, is permissive for tumorigenic growth of HCC with decreasing levels of the cyclin-dependent kinase inhibitor p21
WAF1/CIP1
, suggesting potential utilization of
DCN as an antitumor agent in HCC [
29].
RND3 down-regulation in HCC patients has been reported by several studies [
26,
30,
31], and may be a metastasis suppressor gene in HCC.
However, the expression patterns of four genes among the 20 most significant DEGs in the current study were inconsistent with or ignored in the previous studies, including TBCE, SPINT2, ECM1, and KZAN. The function of KZAN was not identified, whereas the other three genes were all comprehensively studied. In the current study, the inconsistent results might inspire their roles in the oncogenesis and development of HCC with some novel views.
SPINT2 encodes a transmembrane protein with two extracellular Kunitz domains that inhibits a variety of serine proteases. The protein product of
SPINT2 inhibits HGF activator, which prevents the formation of active hepatocyte growth factor, has been taken as a putative tumor suppressor [
32]. Previous studies mainly focus on the methylation of
SPINT2 in HCC instead of its expression [
33,
34]. Nevertheless, we have found that the expression level of
SPINT2 was significantly suppressed in HCC expression profiles. The pattern was consistent with that in cell renal cell carcinoma [
32], which might indicate its potential application as a novel HCC suppressor.
ECM1 encodes a soluble protein that is involved in endochondral bone formation, angiogenesis, and tumor biology. It interacts with a variety of extracellular and structural proteins, contributing to the maintenance of skin integrity and homeostasis [
35]. The expression of
ECM1 is reported to be significantly up-regulated in HCC patients [
24], however, the current analyses of expression profiles showed that expression of
ECM1 was suppressed in HCC patients and were confirmed using RT-PCR. The discrepancy revealed the complicated functions of
ECM1 in the oncogenesis and development of HCC.