Background
Liver cancer is the third leading cause of cancer death and the sixth most common cancer worldwide [
1]. Hepatocellular carcinoma (HCC) accounts for 70–85% of the total liver cancer burden. The accurate diagnosis and proper treatment of HCC is particularly challenging. Options for HCC treatment remain limited, surgical resection is considered the only “curative treatment”. But HCC patients do not have overt symptoms in the early stages, 80% of patients have widespread HCC at the time of diagnosis and are not candidates for surgical treatment. Even with surgical resection, the 5-year survival rate is poor (about 38%) [
2].
Accumulated data from high-throughput analyses by cDNA microarray provide an accurate landscape of gene expression in HCC, and revealed a lot of pathogenic and prognostic genes for HCCs, thus enabled us to delineate some of the key events that might dominate tumor development and progression. Hopefully, translation of this knowledge into new targets and biomarkers might impact HCC decision making, and ultimately improve patient’s outcomes [
3]. However, these studies were performed with different platforms and in different populations, the suggested diagnostic markers and potential therapeutic targets for HCC varied across studies, hence prevented the application of these findings. It’s the high time to merge these cDNA microarray data from different studies using different platforms, to search those widely and stably changed genes, and to find out better diagnostic markers and potential therapeutic targets for HCC.
In the present study, we performed a systematic review of studies that reported cDNA microarray data for both tumor and nontumor liver tissues of HCC patients and came up with a list of five genes that were differentially expressed in tumor and nontumor tissues across different studies and were significantly correlated to HCC prognosis. Among the five genes, except for alanine-glyoxylate and serine-pyruvate aminotransferase (AGXT) which is an essential gene involved in glyoxylate detoxification, the other four genes were well documented in HCC. Here, we reported that AGXT was involved in the progression of HCC and loss of AGXT expression predicted poor prognosis of HCC. AGXT might be a novel diagnostic and prognostic marker and a potential therapeutic target for HCC.
Methods
Data sources and search strategy
Literature retrieval was originally performed with PubMed, a free search engine accessing primarily the MEDLINE database of references and abstracts on life sciences and biomedical topics. The relevant Mesh Terms were chose as the key words, including hepatocellular carcinoma, liver neoplasms, gene expression profiling, human, gene expression regulation, oligonucleotide array sequence analysis. Date of publications was restricted to July 2016. Besides, the literature must be published in English language. In addition to use (((((((carcinoma, hepatocellular[MeSH Terms]) AND liver neoplasms[MeSH Terms]) AND gene expression profiling[MeSH Terms]) AND human[MeSH Terms]) AND gene expression regulation[MeSH Terms]) AND oligonucleotide array sequence analysis[MeSH Terms])) AND (“1781/01/01”[Date-MeSH]: “2016/06/30”[Date-MeSH]) as the search strategy to obtain literatures, we also reviewed the bibliographies of eligible studies as well as those of relevant review articles to identify additional studies not captured by our database searches. To identify all potentially eligible studies, two investigators independently conducted structured searches in selected databases.
Study inclusion criteria
Eligible studies were included in the systematic review if they met the following criteria: (1) they were gene expression studies in hepatocellular carcinoma; (2) they used tissue samples obtained from surgically resected tumor and corresponding non-tumor or normal tissues in human for comparison; (3) validation of method and sample set were reported; (4) clinical and experimental study. Articles were excluded based on the following criteria: (1) review articles or letters; (2) non-liver cancer; (3) non-gene expression profile; (4) non-human tissue samples.
From the full text and corresponding additional information, the following items were eligible to collect and record for each study: authors, year of publication, region, selection number and characteristics of recruited liver cancer patients, the members of abnormally expressed genes in liver cancer. Two investigators (K. D. and YL. J.) independently evaluated and extracted the data with the inclusion criteria. Conflicts in study selection at this stage were resolved by consensus, referring back to the original article in consultation with the principal.
Gene statistics and screening
For genes that have multiple names/synonyms, we standardized the gene name for later work by using widely accepted and used HGNC database (
https://www.genenames.org/). Frequency and composition analysis of the abnormally expressed genes in hepatocellular carcinomas were conducted by IBM SPSS Statistics 20.0 software (Endicott, New York, NY). The genes presented in more than four studies were regarded as high frequent genes. The expression of the high frequent genes in other malignant tumors were obtained from Oncomine (
https://www.oncomine.org/). To identify a prognostic gene list across different studies, the prognostic value of these high frequent genes were evaluated by survival risk prediction in a previously described cohort of 247 Chinese HCC patients with publicly available Affymetrix U133A array data (National Center for Biotechnology Information [NCBI] Gene Expression Omnibus [GEO] Accession number GSE14520) [
4]. BRB-Array Tools (version 4.3.1) was used for survival risk prediction. By reviewing related publications, the prognostic genes that have been intensively investigated were excluded, and the only one gene AGXT (alanine-glyoxylate and serine-pyruvate aminotransferase) that had not been reported in HCC were chosen for further validation.
Tissue microarray and immunohistochemistry staining
Tissue microarrays of 192 HCC patients were used for validation. All of the patients underwent curative hepatectomy for primary HCC at the Affiliated Hospital of Nantong University between March 2004 and August 2009. No patients received either radiotherapy or chemotherapy before the surgery. The study was performed on the basis of the protocol approved by the Declaration of Helsinki, and written informed consent was obtained from all patients. The histological Grading of HCC was defined according to the Edmondson–Steiner grading system. In the present study, grade I and II were termed as well differentiated, grade III was termed as moderate differentiated, and grade IV was termed as poor differentiated. Tumor stage was assigned according to the American Joint Committee on Cancer TNM staging. Patients were followed up every 2–3 months during the 1st year after surgery and every 3–6 months thereafter until September 2016. Totally 101 out of the 192 HCC patients had integrated clinical and follow up data (Additional file
1: Table S1).
Immunohistochemistry was performed with Envision + kits (DAKO, Carpinteria, CA) according to the manufacturer’s instructions [
4], of which positive staining appeared in brown. The primary antibodies used were rabbit anti-AGXT monoclonal antibody (1:100, Abcam). Immunostained slides were analyzed with a semi quantitative scoring approach which combined staining intensity and percentage of positive cells: grade 0 for no reaction or focal weak reaction; grade 1 for intense focal or diffuse weak reaction; grade 2 for moderate diffuse reaction; and grade 3 for intense diffuse reaction, and the corresponding slides were scored from 0 to 3 respectively. The staining scores were evaluated independently by two pathologists who were blinded to the clinical outcomes.
Cell culture
Human HCC cell lines were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (Gibco, CA, USA) supplemented with 10% fetal bovine serum (FBS, Gibco, Carlsbad, USA) in a 5% CO2 atmosphere at 37 °C.
siRNA transfection
Huh-7 cells were seeded into 6-well culture plates and transfected with small interfering RNAs (siRNAs) against human AGXT (siAGXT) (SASI_Hs01_00153951, SASI_Hs01_00153952 and SASI_Hs01_00153953, Sigma) using Lipofectamine™ 2000 (Invitrogen, Carlsbad, CA), at a final concentration of 75 nm. Non-targeting control siRNA (siNC, Sigma) was used as negative control. Cy3 labeled siRNA transfection control (Cy3-siTC, RiboBio, Guangzhou, China) was used to optimize siRNA concentration for transfection. Knockdown efficiency was determined by real-time quantitative reverse transcription-PCR (RT-PCR) 48 h after the transfection.
Real-time quantitative RT-PCR
The RNAs from Huh-7 or HepG2 cells were reverse-transcribed with Thermoscript RT-PCR system (Invitrogen). Real-time quantitative PCR was performed on RotorGene 3000 instrument (Corbett Research, New South Wales, Australia) with SYBR qPCR master lit (TaKaRa, Dalian, China). Gene specific primers were: AGXT forward primer 5′-CTGGGGACTCCTTCCTGGTT-3′, AGXT reverses primer 5′ -CACCTCCTGCAGTGTGTAGT-3′; β-actin forward primer 5′-TTGTTACAGGAAGTCCCTTGCC-3′, β-actin reverses primer 5′ -ATGCTATCACCTCCCCTGTGTG-3′. Relative gene expression was normalized to housekeeping gene β-actin and calculated as 2 − ΔCτ.
Western blot
For western blot analysis, whole cell protein was extracted by RIPA lysis buffer (Beyotime, Shanghai) according to the manufacturer’s instructions. Equal amounts of protein (20 µg) were separated on 10% SDS PAGE gel and transferred onto polyvinyldifluoride (PVDF) membranes (Millipore). PVDF membranes were blocked with 5% non-fat milk for 1 h, then incubated with specific primary antibodies for anti-AGXT antibody (Abcam, Cambridge, UK), and β-actin (CST, Boston, America) at 4 °C overnight, then incubated with horseradish peroxidase-conjugated secondary antibody for an additional 1 h at room temperature. The protein expression was visualized with the ECL chemiluminescence detection system (Biorad).
Proliferation and migration assays
Proliferation ability of HCC cells was measured by cell counting using a Counting Chamber or Cell Counting Kit-8 reagent (CCK8, Dojindo Laboratories, Kumamoto, Japan) according to the manufacturer’s recommendation. Cells were seeded at 1 × 105 cells per 25 cm2 Corning cell culture flask or 2 × 103 cells per well in 96-well plates and cultured for 0 h, 24 h, 48 h, 72 h, 96 h, 120 h and 144 h. For cells cultured in 96-well plates, CCK8 solution was added (10 μl each well) and incubated at 37 °C for 2 h. The optical density readings at 450 nm were determined by a micro plate reader (Bio-Rad, Tokyo, Japan).
For transwell migration assays, a single cell suspension of 50,000 Huh-7 and HepG2 cells, or siAGXT/siNC treated Huh-7 cells resuspended in 200 μl serum-free DMEM were plated in the upper chambers (Millicell, 8.0 μm; Corning, USA), 750 μl DMEM medium with 10% FBS was used as a chemoattractant in the lower chambers. After 24 h, nonmigrating cells were removed from the upper surface softly by a cotton swab. The cells that migrated through the membrane to the lower surface were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet, then counted under a microscope (Olympus) at 200-fold magnification, five fields were counted for each well.
Cell cycle assay and apoptosis assay
HCC cells were seeded at 2 × 105 per well in 6 well plates. After overnight incubation, cells were treated with siAGXT or siNC for 72 h. After treatment, HCC cells were digested using 0.25% pancreatic enzyme, washed twice with PBS, fixed in precooled 70% cold ethanol at 4 °C for 24 h. Cells were centrifuged again, washed with cold PBS twice, and stained with Propidium iodide (0.1 mg/ml) (Propidium iodide, PI; Beyotime) at 37 °C in the dark for 30 min. DNA contents were measured with a BD FACS Calibur system (BD Biosciences, Franklin Lake, NJ). Data were analyzed using ModFit LTTM software (Verity Software House, Topsham, ME). For apoptosis assay, cells were harvested, washed, and resuspended with PBS and stained with BD Annexin V-FITC Apoptosis Detection Kit. Data was acquired using a BD FACS Calibur system and BD FACSuite software (BD Biosciences).
Statistical analysis
Statistical analyses were performed with Graphpad Prism 6.0 (Graphpad Software, Inc., La Jolla, CA) or IBM SPSS Statistics 20.0 software (Endicott, New York, NY). Quantitative data were expressed as mean ± SD. Comparisons between groups were made by Student’s t-test or two-way ANOVA. Categorical data were evaluated by the χ2 test. All P-values were two-sided and the statistical significance was defined as P < 0.05. All experiments were performed in triplicates.
Discussion
To date, a large proportion of existed systematic reviews or meta studies on HCC markers were about serum markers [
67,
68], and others were focused on one or a set of genes [
69,
70], inflammatory-based markers [
71,
72]. Due to the differences across experimental methods, sample size and quality, inconsistent annotation and the methods used for data processing and analysis, it is difficult to do integrated bioinformatics analysis with microarray data from different platforms, and sometimes even the raw data are not available in public databases, so there are very few studies that combined the data form different HCC gene expression profiles [
73,
74]. Zhang et al. reanalyzed three publicly available datasets of gene expression profiles in the Oncomine database; they identified 17 hub genes (10 unregulated and 7 down-regulated in HCC tissues). Among the 17 genes, 13 genes (SMAD2, PTK2, MAPK1, HDAC1, CDC25A, IGFI, FOS, ESR1, EGFR, SOCS3, SP1, YY1 and JUNB) have been identified as an HCC-related gene [
73]. Shi et al. analyzed the integrated microarray data from four independent studies (GSE14520, GSE25097, GSE36376 and GSE57957) from public databases Gene Expression Omnibus (GEO,
http://www.ncbi.nlm.nih.gov/geo); they found that KLHL21 was a potential target for therapeutic intervention [
74].
In the present study, we conducted a systematic review of studies that reported cDNA microarray data and differentially expressed genes between HCC tumor and nontumor tissues. We searched the PubMed databases for eligible studies published in English-language before July 2016 and retrieved 392 articles. The list of differentially expressed genes from 43 carefully designed studies that satisfied further inclusion criteria were summarized. A total of 1917 HCC patients were involved and 2022 non redundant abnormally expressed genes in HCC were extracted. The frequencies of reported genes were ranked. We finally obtained a list of only five genes (AGXT; ALDOB; CYP2E1; IGFBP3; TOP2A) that were differentially expressed in tumor and nontumor tissues across studies, and were significantly correlated to HCC prognosis. Among the five genes, four (ALDOB; CYP2E1; IGFBP3; TOP2A) were well documented HCC related genes, while AGXT was mostly studies in primary hyperoxaluria type 1 [
57,
58], but had not been reported in HCC.
Unlike the unstable changes of the other four genes, the expression of AGXT was constantly reduced in tumors compared to nontumor tissues in a variety of malignant, including liver cancer, gastric cancer, kidney cancer, lung cancer and pancreatic cancer (Additional file
3: Table S3). In the present study, we found that the expression of AGXT reflected the differentiation of HCC and reduced AGXT expression was correlated to poor overall survival of HCC patients. Knocking down of AGXT in HCC cell line could induce a cell cycle shift from G0/G1 to S and G2/M together with enhanced cell proliferation, increased cell death and migration, suggesting a role of AGXT in promoting tumor progression. We provided a high level of evidence on AGXT to serve as a new biomarker and prognostic factor that related to tumor differentiation and progression for HCC.
AGXT gene encoded an enzyme called alanine-glyoxylate and serine-pyruvate aminotransferase which is found in liver cells, specifically within peroxisomes. Serine-pyruvate aminotransferase converts glyoxylate to glycine and is involved in glyoxylate detoxification. The role of AGXT in tumor biology has not been reported yet. One relevant observation is that the metabolic changes to pyruvate and glyoxylate might contribute to the improved therapeutic effects of sorafenib and everolimus combination therapy for HCC [
75]. The underlying mechanisms for the contribution of decreased AGXT expression to HCC progression deserved further investigation.
As we were preparing the manuscript, another study group reported AGXT as a novel immunohistochemical marker for the diagnosis of HCC, and demonstrated comparable specificity and higher sensitivity of AGXT compared to arginase-1 [
76]. The reason why they were interested in AGXT was uninterpreted; they only mentioned that AGXT was expressed in the liver exclusively. Their finding reaffirmed that AGXT as a new HCC biomarker, and backed up the feasibility and necessity of systematic review on discovering new and reliable biomarkers for HCC as well as for other cancer.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.