Identification of key genes in hepatocellular carcinoma and validation of the candidate gene, cdc25a, using gene set enrichment analysis, meta-analysis and cross-species comparison.

The aim of the present study was to determine key pathways and genes involved in the pathogenesis of hepatocellular carcinoma (HCC) through bioinformatic analyses of HCC microarray data based on cross-species comparison. Microarray data of gene expression in HCC in different species were analyzed using gene set enrichment analysis (GSEA) and meta-analysis. Reverse transcription-quantitative polymerase chain reaction and western blotting were performed to determine the mRNA and protein expression levels of cdc25a, one of the identified candidate genes, in human, rat and tree shrew samples. The cell cycle pathway had the largest overlap between the GSEA and meta-analysis. Meta-analyses showed that 25 genes, including cdc25a, in the cell cycle pathway were differentially expressed. Cdc25a mRNA levels in HCC tissues were higher than those in normal liver tissues in humans, rats and tree shrews, and the expression level of cdc25a in HCC tissues was higher than in corresponding paraneoplastic tissues in humans and rats. In human HCC tissues, the cdc25a mRNA level was significantly correlated with clinical stage, portal vein tumor thrombosis and extrahepatic metastasis. Western blotting showed that, cdc25a protein levels were significantly upregulated in HCC tissues in humans, rats and tree shrews. In conclusion, GSEA and meta-analysis can be combined to identify key molecules and pathways involved in HCC. This study demonstrated that the cell cycle pathway and the cdc25a gene may be crucial in the pathogenesis and progression of HCC.

Introduction

Hepatocellular carcinoma (HCC) is one of the most common types of malignancy worldwide (1). It is associated with a poor prognosis in advanced stages. Identifying therapeutic targets for the treatment of HCC is important. During tumor development, somatically acquired “passenger” mutations owing to the inherent genomic instability of cancer cells, gene linkage, and spontaneous mutagenesis may not confer a selective advantage to the developing tumor (2). Currently, one of the challenges in cancer research is identifying key molecular changes among all acquired “passenger” mutations that promote the formation and development of the tumor. In recent years, genome sequencing projects have been completed in a number of animals, including humans, rats and mice, and genes have been identified that have the same expression in different species indicating that these genes could have conservative and important functions. The strategy of cross-species comparative oncogenomics was developed based on this understanding. Mattison et al (3) adopted cross-species comparative genomic hybridization to search for genes that were co-expressed in HCC tissues collected from humans, mice and rats, in order to identify novel candidate genes. The authors of the present study hypothesized that a search for genetic regulators common to humans and other animals, during HCC formation may aid in identifying key genes that affect the pathogenesis and progression of HCC. Gene microarrays have been widely used in HCC research. Analyses on whole-genome mRNA expression microarrays can aid in predicting transcripts that affect the prognosis and recurrence of HCC. However, identifying specific genetic markers that can be used as treatment targets remains a challenge. Mootha et al (4) developed GSEA, with which disease-associated gene pathways can be identified at the genomic level using case-control data. In GSEA, gene expression hybridization data in two biological conditions are analyzed to determine a pattern of gene expression in specific functional gene sets and whether such a pattern is statistically significant. In addition, due to differences in experimental platforms, samples, standardization methods and analytical methods, microarray data obtained in different laboratories may differ. Meta-analysis can be a viable solution to this problem, as it can be used to collect and quantitatively analyze data published on the same subject in an integrative manner, thus obtaining more accurate or a larger number of results than can be gained from any individual study (5). In the present study, GSEA and meta-analysis were combined to analyze whole genome and microarray data from five HCC data sets.

Materials and methods

Databases

A systematic literature and database search was performed to identify the HCC-related gene expression profiles of humans and other animals. Relevant data were downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). The keyword used for the search was 'hepatocellular carcinoma' and the research type was set as expression profiling by array. A total of 230 studies have published gene microarray data. The data sets that met the following criteria were included in this study: The data set must contain whole-genome mRNA expression microarray data; data included a comparison between HCC and normal tissues; both standardized and raw data sets were examined; and the data set had to include >3 samples. Using the above criteria, only five data sets (6–10) were included in the present study (Table I). GSEA and meta-analysis were combined to analyze whole genome and microarray data of these five HCC data sets. The genes that showed significantly differential expression were compared with the mRNA microarray results of a study conducted by our group using the tree shrew HCC model (11) to identify genes in HCC tissue that showed specific changes in >2 species (including humans).

Table I

Basic information on the five whole-genome data sets.

Data set	Authors, (ref.)	Microarray platform	Experimental design	Number of probes	Species	Sample (n)	Control (n)
GSE19665	Deng et al (6)	U133 Plus 2.0	Paired tissues	54000	Homo sapiens	10	10
GSE9809	Liao et al (7)	Mouse430_2	Unpaired+paired tissues	45000	Mus musculus	3	7
GSE9012	Khetchoumian et al (8)	Mouse430_2	Unpaired tissues	45000	Mus musculus	5	5
GSE19004	Viatour et al (9)	Mouse430_2	Unpaired tissues	45000	Mus musculus	5	4
GSE2127	Sheth et al (10)	moe430a	Paired tissues	22000	Mus musculus	9	6

Paired, control for the HCC group came from the same HCC individuals; unpaired, control for the HCC group came from normal individuals.

Sample collection

Human liver tissue samples

All patients have provided written informed consent to have their tissues stored and used for future research. The ethics committee of The Affiliated Tumor Hospital of Guangxi Medical University (Guangxi, China) approved this study. A total of 38 HCC tissue samples and corresponding paraneoplastic tissue samples (>2 cm away from the edge of the tumor) were obtained from patients (age, 24–70 years; mean age, 49.5±11.7) who underwent surgical resection at the Affiliated Tumor Hospital of Guangxi Medical University between January 2009 and December 2011. The diagnosis of HCC in each patient was confirmed histopathologically. Variables including tumor stage, presence of portal vein tumor thrombosis, number and size of tumors, serum α-fetoprotein, tumor differentiation and presence of extrahepatic metastases and recurrence were evaluated in the patients with HCC. The tumor stage was determined according to the Barcelona-Clinic Liver Cancer staging classification (12). Tumor differentiation was graded according to the Edmondson grading system (12). Prior to surgical resection, none of the patients with HCC received chemotherapy. In addition, ten normal liver tissue samples surgically resected from patients with benign liver lesions were also collected.

Rat and tree shrew liver tissue samples

Animal experiments were conducted in accordance with the guidelines for The Care and Use of Laboratory Animals issued by the Ethics Committee of the Affiliated Tumor Hospital of Guangxi Medical University. Rat and tree shrew animal models of HCC induced by aflatoxin B1 (AFB1) have been previously established by our group (13,14). A total of 42 female Wistar rats (age, 4 weeks; weight, 180–210 g) were housed individually under controlled light (12 h light/dark cycle) and temperature conditions. Food and water were available ad libitum. The rats were randomly divided into two groups: An AFB1 group (n=28) and a control group (n=14). The rats in the AFB1 group were intraperitoneally injected with AFB1 (100–200 µg/kg) 1–3 times/week. Liver biopsies were obtained from all rats during the 14 th, 28 th, 42 nd and 55 th week, and all rats were sacrificed on the 64 th week by cervical dislocation. A total of 15 HCC tissue samples and 15 corresponding paraneoplastic tissues samples were collected from the rats in the AFB1 group. A total of 14 normal liver tissue samples were collected from the rats in the control group. Adult tree shrews (n=27; age, 6 months; weight, 100–160 g) were obtained from the Kunming Institute of Zoology, Chinese Science Academy (Yunnan, China), and were allowed to acclimatize for one week to the facilities prior to the experiment. The tree shrews were housed in individual, suspended, stainless steel wire cages, under controlled environmental conditions with a 12 h light/dark photoperiod. They had ad libitum access to tap water and a natural diet containing rice powder (20%), corn powder (20%), beef (20%), wheat bran (10%), soy bean (10%), egg (10%), whole milk powder (5%), and a sugar, salt, vitamin, mineral mix (5%). They were also fed daily with reconstituted powdered milk and fruit. The shrews were randomly divided into an AFB1 group (n=15; 9 males and 6 females) and a control group (n=12; 6 males and 6 females), and as previously reported (15), the animals in the AFB1 group were fed AFB1 (400 mg/kg body weight/day in a small amount of milk; Sigma-Aldrich, St. Louis, MO, USA) from the 1 st to the 90 th week of the experiment, whereas the animals in control group were raised without AFB1 administration. Liver biopsies from each animal were obtained every three months under Ketamine Hydrochloride anesthesia (50 mg/kg; Fujian Gutian Pharmaceutical Co., Ltd., Fujian, China). In the present study, 10 HCC tissue samples and 10 corresponding paraneoplastic tissue samples were obtained when the animals developed cancer (HCC appeared at the 105 th week of experimentation) following sacrifice by cervical dislocation. A total of 10 normal liver tissue samples were collected from tree shrews in control group. After surgical resection, all tissue samples were subjected to rapid freezing in liquid nitrogen and then stored at −80°C. HCC samples were confirmed histopathologically.

GSEA and meta-analysis

Bioconductor (16) 2.10.1 was used to standardize the data. The Robust Multi-array Average (RMA) algorithm (17,18) in the software package, Affy, was used for background correction, standardization and Log2 conversion of raw data on the Affymetrix platform. A t-test was performed to evaluate each probe in each data set. Only genes that were included in the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (19) were subjected to GSEA. Genes with an interquartile range (IQR) <0.5 were excluded. If one gene corresponded to several probes, only the probe with the highest IQR was used. GSEA was performed using the Category package in Bioconductor. The gene sets represented by ≥10 genes were subjected to further analyses, and genes in each pathway were subjected to a t-test. After 1,000 permutations, the P-value of each pathway was obtained. The upregulated and downregulated pathways in all five data sets were compared. The cell cycle pathway was upregulated in all five data sets. All genes in the cell cycle pathway were subjected to meta-analysis in each data set. T-tests were conducted using SAS9.13 software (Cary, NC, USA) to calculate the P-value of each probe in the cell cycle pathway in each data set and the χ2 value of each gene was calculated using the following formula:

In which the degree of freedom is two times that of the data set K). Genes with P<0.05 were obtained. Analyses on these gene pathways were performed using the Database for Annotation, Visualization and Integrated Discover (DAVID; http://david.abcc.ncifcrf.gov/) in KEGG.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)

Total RNA was extracted from tissue samples using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Waltham, MA, USA), and first-strand cDNA was synthesized by reverse transcription from 1 µg RNA using a RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific, Inc.). RT-qPCR was performed using a standard protocol from the SYBR® Premix Ex Taq kit (Takara, Dalian, China) on an Applied Biosystems 7300/7500 Real Time PCR system (Applied Biosystems, Grand Island, NY, USA). The cycling conditions were as follows: 95°C for 30 sec, and 40 cycles at 95°C for 5 sec followed by 60°C for 34 sec. The relative mRNA expression levels of cdc25a to control GAPDH were analyzed using ABI PRISM 7300 software v1.3.1 (Applied Biosystems) and calculated with the double standard curves method. The cdc25a forward and reverse primer sequences were 5′-CCAAAGGAACCATTGAGAAC-3′ and 5′-CAGATGCCATAATTTCTGGAG-3′, respectively, and the product length was 138 bp. The forward and reverse primer sequences for the internal control gene, 3-phosphate dehydrogenase (GAPDH), were 5′-AAGAAGGTGGTGAAGCAGGC-3′ and 5′-ACCACCCTGTTGCTGTAGCC-3′, respectively, and the product length was 200 bp.

Western blot analysis

Western blot analysis to assess cdc25a and GAPDH expression was performed as previously described (14). Protein samples (60 µg) were separated by 10% SDS-PAGE, prior to being transferred onto PVDF membranes (EMD Millipore, Billerica, MA, USA). The membranes were subsequently blocked with Tris-buffered saline with Tween 20 (TBST; Beyotime Institute of Biotechnology, Haimen, China) containing 5% (w/v) skimmed milk powder for c. 2 h at room temperature and then incubated at 4°C overnight with the following primary antibodies: Rabbit polyclonal anti-human Cdc25 (cat. no. ab75743; Abcam, Cambridge, MA, USA) or mouse monoclonal anti-human glyceraldehyde 3-phosphate dehydrogenase (GAPDH; cat. no. TA-08; Beijing Zhongshan Golden Bridge Biotech Co., Ltd., Beijing, China) at dilutions of 1:100 or 1:1000, respectively. The membranes were then washed three times with TBST buffer, and incubated for 1 h at room temperature with IRDye 680LT goat anti-mouse secondary antibody (cat. no. 0926-68020; LI COR Biosciences, Lincoln, NE, USA) or IRDye 680LT goat anti-rabbit secondary antibody (cat. no. 926-68021; LI-COR Biosciences) at 1:7,500 dilution. The membranes were scanned using an Odyssey infrared imaging system (LI-COR Biosciences). Quantification of bands was achieved by ratiometric analysis of the fluorescent intensities of cdc25a and GAPDH using Odyssey Application Software 3.0 (LI-COR Biosciences).

Statistical analysis

SPSS 13.0 (SPSS, Inc., Chicago, IL, USA) was used for data analysis. Quantitative data are expressed as the mean ± standard deviation. Comparison of mean values between multiple groups was performed using a one-way analysis of variance. P<0.05 was considered to indicate a statistically significant difference.

Results

Results of GSEA analysis

GSEA was performed for functional gene enrichment of the five data sets to search for key upregulated and downregulated pathways affecting these data sets. In GSE19665, 27 upregulated and 71 downregulated pathways were enriched; in GSE9809, 56 upregulated and 4 downregulated pathways were enriched; in GSE9012, 66 upregulated and 34 downregulated pathways were enriched; in GSE19004, 73 upregulated and 51 downregulated pathways were enriched; and in GSE2127, 69 upregulated and 50 down-regulated pathways were enriched. There was a large amount of overlap between pathways in the GSE19665 and GSE19004 data sets. Downregulated pathways common to all five data sets included the linoleic acid metabolic pathway and the arachidonic acid metabolic pathway. Upregulated pathways common to all five data sets included the amino sugar and nucleotide sugar metabolic pathways, the cell cycle pathway and the thyroid cancer pathway.

Results of meta-analysis

A total of 220 positive genes were found in the GSE19665 data set and 213 positive genes were found in the remaining four data sets. After the P-value of each gene was obtained, SAS13.0 software was used for meta-analysis. A total of 1,708 genes were found to have P<0.05 and 24 genes were found to have P<10−4 (Table II).

Table II

Distribution of differentially expressed genes revealed by meta-analyses of five data sets.

Gene	P-value
ABI3BP	3.23E-06
CCNB1	5.37E-06
NEK2	7.66E-06
MKI67	1.35E-05
cdc20	2.46E-05
angptl6	2.47E-05
rrm2	2.74E-05
Ttc36	3.06E-05
UBE2C	3.18E-05
mcm2	4.07E-05
ASPM	4.29E-05
NCAPH	5.23E-05
TUBA1B	5.23E-05
CCNB2	5.45E-05
Hist1h2ad	5.71E-05
TOP2A	5.91E-05
FOXM1	6.58E-05
BIRC5	7.18E-05
STMN1	7.74E-05
racgap1	7.84E-05
Hist1h2ag	7.94E-05
Hist1h2ah	7.94E-05
Hist1h2ai	7.94E-05
CDCA5	9.89E-05

The 1,708 genes were subjected to pathway enrichment using the DAVID KEGG database. A total of 720 of these genes were found in the KEGG database.

Overlapping results obtained with GSEA and meta-analysis

The cell cycle pathway had the largest overlap between the GSEA and meta-analysis. Gene probe numbers in the cell cycle pathway of the five data sets were obtained using R programming language. The probe numbers were sent to the website, http://david.abcc.ncifcrf.gov/conversion.jsp, to obtain the official names of the genes. There were 99 differentially expressed genes in the cell cycle pathway of data set GSE19665, 96 in GSE9809, 90 in GSE9012, 106 in GSE19004 and 113 in GSE2127. Meta-analyses demonstrated that 25 genes involved in the cell cycle pathway were differentially expressed (P<0.05). The names, χ2 values and P-values of these genes are shown in Table III.

Table III

Distribution of differentially expressed genes in the cell cycle pathway.

Gene	χ2 value	P-value
CCNB2	46.94	9.67E-07
mcm2	36.51	6.89E-05
YWHAB	35.63	9.74E-05
CCNA2	33.83	2.00E-04
CDKN2C	32.60	3.00E-04
Cdk1	32.24	4.00E-04
MCM6	32.07	4.00E-04
cdkn2b	30.15	8.00E-04
CCNB1	30.15	8.00E-04
CDC25A	29.98	9.00E-04
Mad2l1	29.43	1.10E-03
MCM7	28.57	1.50E-03
CCNE1	28.21	1.70E-03
MCM4	46.94	3.00E-03
cdc20	36.51	6.50E-03
smc3	35.63	6.70E-03
pcnA	33.83	7.90E-03
RAD21	32.60	9.00E-03
CDKN1A	32.24	9.60E-03
CCND1	32.07	1.08E-02
SMAD3	30.15	1.16E-02
TGFB1	30.15	1.27E-02
YWHAZ	29.98	2.06E-02
YWHAG	29.43	2.08E-02
YWHAH	28.57	2.52E-02

Cdc25a mRNA expression

The results of the present study identified 25 differential expression genes in the cell cycle signaling pathway, as determined by GSEA and meta-analysis which analyzed five data sets from human HCC tissue samples. Among the 25 candidate genes, Cdc25a was also shown to be overexpressed in the HCC tissue samples of diethylnitrosamine-induced rats and aflatoxin B1-induced tree shrews (11,20). Therefore, Cdc25a mRNA expression was further investigated. Results of real-time fluorescent PCR showed that the cdc25a mRNA expression level in human HCC tissue (0.00425±0.00241) was significantly higher than in the corresponding paraneoplastic tissue (0.00086±0.00081) (P<0.05) and in normal liver tissue (0.00038±0.00032) (P<0.05). There was no significant difference between the paraneoplastic tissue and normal liver tissue (P>0.05). The cdc25a mRNA expression level in rat HCC tissue (0.00281±0.00278) was significantly higher than that in the corresponding paraneoplastic tissue (0.00044±0.00035) (P<0.05) and in normal liver tissue (0.00051±0.00022) (P<0.05); however, there was no significant difference between the paraneoplastic tissue and normal liver tissue (P>0.05). The cdc25a mRNA expression level in tree shrew HCC tissue (0.00043±0.00021) was significantly higher than that in normal liver tissue (0.00022±0.00010) (P<0.05), however there was no significant difference between the paraneoplastic tissue (0.00028±0.00017) and corresponding HCC tissue or normal liver tissue (P>0.05; Fig. 1).

Figure 1

Expression of cdc25a mRNA in the HCC, paraneoplastic and normal liver tissues of different species (humans, rats and tree shrews). HUM, human; RAT, rat; TS, tree shrew.

The 38 HCC patients were grouped on the basis of whether they had PVTT, extrahepatic metastasis, BCLC stage, postoperative recurrence, tumor diameter, number of tumors, serum AFP level and degree of tumor differentiation. The correlation between cdc25a mRNA expression in the HCC tissues and the above clinical parameters was examined. As demonstrated in Table IV, the detection rate of cdc25a mRNA was observed to be significantly correlated with Barcelona Clinic Liver Cancer (BCLC) stage, portal vein tumor thrombus (PVTT) and extrahepatic metastasis, but not with postoperative recurrence, tumor diameter, tumor number, serum α-fetoprotein (AFP) level or degree of tumor differentiation.

Table IV

Correlation between the detection rate of cdc25a mRNA in human HCC tissues and clinical parameters.

Clinical parameter	n	Cdc25a mRNA expression level	t-value	P-value
BCLC stage (12)
0, A	26	0.00253±0.00175	2.342	0.023
B, C	12	0.00537±0.00221
PVTT
Yes	10	0.00564±0.00259	0.505	0.030
No	28	0.00376±0.00214
Extrahepatic metastasis
Yes	11	0.00571±0.00254	0.139	0.014
No	27	0.00367±0.00208
Recurrence
Yes	20	0.00408±0.00250	0.071	0.632
No	18	0.00446±0.00229
Tumor diameter (cm)
≥5	31	0.00436±0.00236	0.218	0.568
<5	7	0.00378±0.00262
Number of tumors
1	24	0.00375±0.00203	1.886	0.083
≥2	14	0.00513±0.00274
Serum AFP (µg/l)
≥400	13	0.00521±0.00249	0.215	0.075
<400	25	0.00376±0.00221
Tumor differentiation
Highly differentiated	21	0.00434±0.00217	1.602	0.825
Poorly differentiated and undifferentiated	17	0.00416±0.00268

BCLC, Barcelona Clinic Liver Cancer; PVTT, portal vein tumor thrombus; AFP, α-fetoprotein.

Cdc25a protein expression

Based on the results of fluorescent quantitative PCR, 38 human HCC and corresponding paraneoplastic tissue samples, 10 normal human liver tissue samples, 15 rat HCC and corresponding paraneoplastic tissue samples, 10 normal rat liver tissues, 10 tree shrew HCC and corresponding paraneoplastic tissue samples, and 10 normal tree shrew liver tissue samples were collected for western blot analysis. Specific cdc25a and GAPDH protein bands were observed at 36 and 59 kDa of pre-stained proteins, respectively, although the target protein, cdc25a, was not detected in normal human liver tissue, or in the normal or paraneoplastic liver tissues of rats or tree shrews. The mean relative cdc25a expression level in human HCC tissues (0.339±0.239) was significantly higher than in the corresponding paraneoplastic tissues (0.0609±0.0498; P<0.05). In humans, rats and tree shrews, the cdc25a bands were notably stronger in HCC tissues than in the corresponding paraneoplastic and normal liver tissues, consistent with the high cdc25a mRNA expression in HCC tissues (Fig. 2).

Figure 2

Cdc25a protein expression in the HCC, paraneoplastic and normal liver tissues of different species (humans, rats and tree shrews), revealed by western blot analysis. (A) cdc25a protein expression in human liver tissues; (B) cdc25a protein expression in rat liver tissues; and (C) cdc25a protein expression in tree shrew liver tissues. Lanes 1 and 3, HCC tissue; lanes 2 and 4, corresponding paraneoplastic tissue; and lanes 5 and 6, normal liver tissue. GAPDH, 3-glyceraldehyde phosphate dehydrogenase.

Discussion

Analyses on gene microarray data are an important part of gene microarray research. A lot of useful information may be missed due to potential problems with the samples when analyzing the results of a single experiment. In addition, using a t-test to target a single set of microarray data has certain limitations. The small sample size may lead to an unreliable variance estimation, resulting in a high false-positive rate, whereas differences in the expression levels of different samples may be ignored (21). In GSEA, microarray data of samples in two different biological states (e.g., normal vs. cancerous) are analyzed to determine whether a set of genes show similar expression trends between the two states, thereby identifying genes or pathways associated with the disease (22).

In the present study, GSEA and meta-analysis were combined to analyze five data sets, and the two results were compared in order to identify HCC-related genes and pathways. DEGs in >2 groups of samples were analyzed using GSEA and the clustering was performed using differentially expressed genes. R programming language was used to process the data and for statistical analysis to obtain pathways that showed changes common to all five data sets. DEGs were analyzed using the DAVID website to determine the pathways in which these genes may be involved. The cell cycle pathway statistically significant in the results obtained with the GSEA and meta-analysis and was further analyzed to identify genes with significant differential expression.

Among the 25 differentially expressed genes in the cell cycle pathway that were identified, cyclin B2, cyclin B1, cyclin D1, cdc25a and cdk1 are closely associated with HCC occurrence and development (23–26). Cyclin-dependent kinases (CDKs) are core proteins in the cell cycle regulatory network. Changes in the expression activity of CDKs directly affect the length of the cell cycle, determine the progression of the cell cycle, and are closely associated with the growth, differentiation, migration and apoptosis of normal cells, as well as the occurrence, development and metastasis of tumors (27). CDKs are important in the regulation of hepatoma cell proliferation and apoptosis. Cyclin B and cyclin D are members of the cyclin family and determine which CDKs phosphorylate which substrates as well as when and where. Studies have shown that (27) Cyclin D1 is expressed in normal liver tissue, but its expression is increased in HCC tissue, and is correlated with the histological grade of HCC. This suggests that Cyclin D1 is involved in the progression and development of HCC and may promote cell proliferation, contributing to tumor formation. Cyclin B1 is in the same family, is highly expressed in HCC tissue and is key in the transition process of HCC cells. Cdc25A is a phosphatase with dual specificity, and can activate CDKs, promoting the progression of the cell cycle.

Studies have shown that cdc25a is associated with breast cancer (28), non-small cell lung cancer (29), colorectal cancer (30), prostate cancer (31) and other malignant tumors. However, there are few studies regarding the correlation between cdc25a and liver cancer. In a study by Xu et al (26), cdc25a expression was examined in HCC tissues using RT-PCR, immunohistochemistry and western blotting, and reported cdc25a expression detection rates of 69 (9/13), 56 (33/59) and 78% (46/59), respectively. Immunohistochemistry tests showed that high cdc25a protein expression was positively correlated with poor tumor differentiation and PVTT. In the present study, fluorescent quantitative PCR and western blotting showed that cdc25a expression was high in human HCC tissues at the transcriptional and at the translational levels. The cdc25a mRNA expression level in human HCC tissues was found to be positively correlated with PVTT and extrahepatic metastasis. This suggests that cdc25a is associated with the progression of HCC, and may serve as an indicator to evaluate the severity of the disease. The differences between the results of the present study and those of Xu et al (26) may be due to differences in sample sources, transcription and translation levels, and experimental method.

In this study, overexpression of cdc25a was also found in HCC samples from aflatoxin B1-induced rat and tree shrew HCC models. Such consistent differential expression in different species has also been reported in a previous study by our group (11). The same cross-species expression patterns suggest that cdc25a is key in the pathogenesis and progression of HCC. Xu et al (32) explored the possibility of using cdc25a as an antitumor target by inhibiting its activity in HCC cell lines. It was reported that antisense oligonucleotides of cdc25a inhibited 25–50% of cell growth within 48 h, resulting in arrest at the G0–G1 phase and effectively suppressing the proliferation of HCC cells. This suggests that cdc25a can serve as a feasible target for anticancer treatment. In addition, a study by Liu et al (33) showed that knockdown of the ROCK2 gene could activate the reduced ubiquitin-proteasome pathway, thereby promoting cdc25a ubiquitination, which would ultimately lead to cdc25a degradation and inhibition of HCC cell growth. Therefore, cdc25a may be a novel target for anti-HCC treatment.

In conclusion, GSEA and meta-analysis can be combined to identify key molecules and pathways involved in the pathogenesis and progression of HCC. The cell cycle pathway and the cdc25a gene may be the crucial in the pathogenesis and progression of HCC.