Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Early diagnosis and prognosis of hepatocellular carcinoma based on a ceRNA array

Early diagnosis and prognosis of hepatocellular carcinoma based on a ceRNA array IntroductionHepatocellular carcinoma (HCC) has one of the highest incidences and mortality rates worldwide. It involves the carcinogenesis of liver parenchymal cells. In 2020, global cancer data showed 910,000 new liver cancer cases and 830,000 deaths, both of which were in the top 10 of all cancers [1]. Many factors such as infection (HBV, HCV, and parasites), eating food containing aflatoxin, and heavy drinking are related to the occurrence of liver cancer. Despite the development of effective anti-viral drugs, the incidence of HCC is increasing worldwide [2]. Although surgical resection, hepatic artery embolization, radiofrequency ablation, radiotherapy, and chemotherapy have significantly increased the overall survival (OS) of patients with HCC, the prognosis remains very poor, due to a lack of early diagnoses [3, 4]. Therefore, early diagnostic markers and therapeutic targets for HCC need to be urgently explored.Competing endogenous RNA (ceRNA) networks combine the functions of protein coding and non-coding RNAs (ncRNAs), and their mutual communication and co-regulation have important roles in human development and disease processes [5, 6]. The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO), and other public databases are indispensable for bioinformatics analysis, from which researchers can download tumor gene expression profiles for weighted gene co-expression analysis to construct ceRNA networks, which provide important information for tumor marker screening and pathogenic mechanisms [7]. For example, a study found that a ceRNA network composed of three lncRNAs (LINC01106, FOXD2-AS1, and AC103702.2) and three mRNAs (CCDC34, ORC6, and Sox4) was strongly associated with the prognosis of gastric cancer [8]. Three lncRNAs (Dirc3, DNAH10OS, and GK-IT1) were screened from the constructed ceRNA network and were used for evaluation of esophageal cancer risk models [9].Based on HCC information from the TCGA database, we established a lncRNA-miRNA-mRNA ceRNA network to explore the key genes related to the diagnosis and prognosis of HCC, and verified their expression levels and the targeted binding relationship between them in clinical tissues.Materials and methodsAccess to ceRNA network genesThe gene expression profiling information of HCC was obtained from the TCGA database, and differential gene screening was completed using the DESEQ2 package in R4.0.2 software [false discovery rate, (FDR)<0.05; | Log(fold change) |>1], mainly targeting middle lncRNA, miRNA, and mRNA.Construction of the ceRNA networkInteractions between lncRNA, miRNA, and mRNA were analyzed using miRcode [10] and starbase [11] databases (hypergeometric testing, p<0.05; correlation testing, p<0.05). Cytoscape 3.7.2 was used to establish the ceRNA network relationship.The DAVID database was used to perform Gene Ontology and KEGG analyses (p<0.01).Identification of targeted genes of the key miRNATo obtain more reliable targets of core miRNA, TargetScan [12], miRDB [13], and PicTar [14] were searched to identify the intersections with starBase.Validation by qRT-PCR in clinical tissuesBased on the results of bioinformatics analysis, one lncRNA (FOXD2-AS1), one miRNA (has-miR-9-5p), and three mRNAs (STMN1, COL15A1, CCNE2) were identified as candidate genes. Clinical tissue chips were used for gene expression analysis [chip lot numbers: cDNA-HLivH30PG02 and cDNA-HLivH087Su02 (Shanghai Outdo Biotech Co., Ltd. Shanghai, China)], containing a total of 96 cancer tissue samples and 51 normal adjacent tissue samples. The tailing reaction was used for reverse-transcription of miRNA, and the primers were universal. The primers are shown in Table 1. ACTB, encoding β-actin was used for internal reference of lncRNA and mRNA, and U6 was applied for miRNA. The experiment was conducted strictly according to the instructions of the PerfectStart™ Green qPCR SuperMix (TransGen Biotech, Beijing, China) and miRcute Plus miRNA qPCR detection kit (SYBR Green) (Tiangen, Beijing, China). Primer synthesis was completed by Sangon Biotech Co., Ltd., and the primer for miR-9-5p was provided by Tiangen Co., Ltd.Table 1:Primers for lncRNA and mRNA.GenePrimer sequence (5′–3′)FOXD2-AS1FGGCACAGCGGAACGATTGRTGTACCGGCTGAGGCTAGSTMN1FTGGTAGAAATACCCAACGCRGAGGCATCCAAACAAAGCAGTCOL15A1FTGCGACTGAAACAAACAGGTTCARGAAGAAAGCACCCGGAAAAGATGCCNE2FGCATTATGACACCACCGAAGARTAGGGCAATCAATCAATCACAGCF, forward; R, reverse.Tissue cDNA samplesAll HCC tissuecDNA was provided by commercial cDNA microarrays, which were purchased from Shanghai Xinchao Biotech Co., Ltd. The batch numbers of the chips were CDNA-HLIVH30PG02 and cDNA-HLivH087Su02, which contained 96 cancer tissue samples and 51 normal adjacent tissue samples.Dual luciferase reporter assayA wild-type 3′-UTR and its mutants were designed for FOXD2-AS1, STMN1, COL15A1, and CCNE2 binding site sequences to miR-9-5p. The 3′-UTR and its mutants were cloned downstream of the luciferase fragment of a pMiR reporter vector, and then the constructed vector was sequenced. The vector and chemically synthesized miR-9-5p fragment (NC mimics as a control group) were co-transfected into 293T cells. After 48 h, luciferase activity was detected by using a dual-luciferase system (Promega, USA). There were two miR-9-5p binding sites in the 3′-UTR sequence of FOXD2-AS1, and therefore, the base fragments of these two binding sites were connected in series to develop the 3′-UTR and mutant of FOXD2-AS1, which was cloned into a pMiR reporter vector for analysis.Statistical analysisThe GDCRNATools, GGPLOT2, DESEQ2, survival, glmnet, and survminer modules in R4.0.2 software were used to analyze HCC expression profile data from TCGA. Single-factor Cox regression and lasso regression analyses were employed to ensure that the model was not overfitting and all significant genes were integrated into the Cox regression model. Risk survival and the ROC curves were used to evaluate the diagnostic value of the model. Kaplan-Meier survival analysis was applied in batches for each gene in the ceRNA network to assess its hazard. The relative expression of key genes in HCC tissues was calculated via qRT-PCR with the 2⁻△△Ct method, and the difference in the relative expression between the HCC cancer group and the para-cancer group was analyzed using an independent sample t-test. The ROC curve was drawn according to qRT-PCR detection of the tissue cDNA chip. The diagnostic value of all important members, ceRNA combinations and candidate ceRNA arrays was evaluated with the AUC value. The statistical analysis and plotting were performed using IBM SPSS Statistics 25.0. and GraphPad Prism 8.ResultsLncRNA-miRNA-mRNA ceRNA in HCCIn this study, HCC transcriptome data from 424 samples (adjacent: 50 cases, cancer: 374 cases), miRNA data from 425 samples (adjacent: 50 cases, cancer: 375 cases), and related clinical information were downloaded from the TCGA database. Survival analysis and univariate analysis were performed on the clinical information of the 366 HCC patients in the TCGA database, which revealed that T stage, M stage and N stage were associated with survival of HCC patients (see Table 2).Table 2:Clinical correlation analysis of TCGA database baseline data.GroupNCumulative survival rateLog-rankUnivariate analysis3 years5 yearsX2p-ValueHR95% CIp-ValueAge, years3661.3840.239≤6017562.1%54.6%1＞6019162.3%47.7%1.2350.868–1.7580.240Gender3661.2040.273Female11858.1%45.3%1Male24865.1%52.9%0.8180.571–1.1720.273Grade3661.2750.735G15567.8%55.2%1G217862.1%50.1%1.1830.699–2.0030.531G312159.2%48.1%1.2380.713–2.1490.448G4 TNM12––1.7120.632–4.6360.290T36639.373≤0.001T117977.1%64.6%1T29559.4%43.2%1.5200.959–2.4100.075T37944.4%33.4%2.7151.775–4.151≤0.001T413005.5052.742–11.053≤0.001N3664.6800.096N025066.4%58.8%1N1433.3%-2.1110.517–8.6270.298NX11253.8%27.4%1.4571.002–2.1190.049M36610.8100.004M026467.4%58.8%1M14004.2631.340–13.5660.014MX9852,6%31.9%1.5471.058–2.2610.024Stage36634.761≤0.001Ⅰ17477.4%65.8%1Ⅱ9557.6%41.9%1.6601.045–2.6390.032Ⅲ8842.0%31.6%2.9221.919–4.450≤0.001Ⅳ914.3%04.6842.097–10.460≤0.001A total of 136 DE-lncRNAs (104 upregulated, 32 downregulated) (Figure 1A), 128 DE-miRNAs (107 upregulated, 21 downregulated) (Figure 1B), and 2028 DE-miRNAs (1222 upregulated, 806 downregulated) (Figure 1C) were obtained using gene differential expression analysis.Figure 1:Competing endogenous RNA-related gene screening. (A) Heat map and volcano plot of differentially expressed lncRNAs. (B) Heat map and volcano plot of differentially expressed miRNAs. (C) Heat map and volcano plot of differentially expressed mRNAs.The differentially expressed lncRNAs (n=8), miRNAs (n=5), and prognostic mRNAs (n=21) were used to construct a ceRNA network (Figure 2A) (see Table 3).Figure 2:CeRNA network and functional enrichment analysis. (A) Competing endogenous RNA network in HCC. (B) Results for KEGG enrichment analysis of the mRNAs in the ceRNA network.Table 3:LncRNA, miRNA and mRNA that constituted the ceRNA network.RNA typeGene IDLncRNAAC092171.2, AC239868.2, CRNDE, PVT1, FOXD2-AS1, AL445524.1, LINC00511, AC111000.4miRNAhsa-miR-96-5p, hsa-miR-182-5p, hsa-miR-424-5p, hsa-miR-217, hsa-miR-9-5pmRNAANLN, KIF23, CCNE1, RBM24, GNAL, SLC6A9, ZIC2, STMN1, CTHRC1, CBX2, DTL, EZH2, CCNE2, APLN, SNCG, GPC3, LRRC1, NRCAM, IGF2BP1, TSPAN5, COL15A1The detailed regulatory relationships of lncRNA-miRNA and miRNA-mRNA are described in Table 4. Functional enrichment analysis of GO and KEGG showed that 21 mRNAs were mainly enriched in 14 signaling pathways, of which “microRNAs in cancer pathway” was the most important (Figure 2B).Table 4:Regulatory relationships between genes in ceRNA network.lncRNAProtein-coding RNAmiRNAHypergeometric test p-ValueCorrelation p-ValueAC092171.2ANLNhsa-miR-96-5p0.0002331.85E-30hsa-miR-424-5phsa-miR-182-5pKIF23hsa-miR-2170.0064132.57E-28hsa-miR-424-5CCNE1hsa-miR-424-5p0.0011791.37E-24RBM24hsa-mi R-424-5p0.003890.000133GNALhsa-miR-424-5p0.0057453.40E-06SLC6A9hsa-miR-96-5p0.0001091.20E-09hsa-miR-182-5pAC239868.2ZIC2hsa-miR-96-5p0.0041161.15E-17STMN1hsa-miR-9-5p0.0074184.86E-21CTHRC1hsa-miR-9-5p0.0005890.0000108CBX2hsa-miR-96-5p0.0001261.97E-12CRNDEDTLhsa-miR-2170.0063071.69E-09EZH2hsa-miR-2170.0003651.58E-11CCNE2hsa-miR-9-5p0.0033760.000215PVT1APLNhsa-miR-424-5p0.0004450.00073KIF23hsa-miR-424-5p0.0001091.24E-06CTHRC1hsa-miR-9-5p0.001371.23E-13SNCGhsa-miR-424-5p0.0004891.41E-19GNALhsa-miR-424-5p0.0036841.15E-08FOXD2-AS1STMN1hsa-miR-9-5p0.0019441.01E-26CCNE1hsa-miR-9-5p0.0077148.93E-29AL445524.1GPC3hsa-miR-96-5p0.0011591.23E-11hsa-miR-182-5pANLNhsa-miR-96-5p0.0011910.000745hsa-miR-182-5phsa-miR-217LRRC1hsa-miR-9-5p0.0003050.000682NRCAMhsa-miR-182-5p0.0042920.0003LINC00511IGF2BP1hsa-miR-9-5p0.0003939.07E-17CTHRC1hsa-miR-9-5p0.0001681.84E-19CBX2hsa-miR-424-5p0.000149.90E-16SNCGhsa-miR-424-5p0.0017472.04E-08GNALhsa-miR-424-5p0.0017471.27E-23TSPAN5hsa-miR-424-5p0.0008830.004075AC111000.4COL15A1hsa-miR-9-5p0.0009095.40E-05Determination of the ceRNA arraySingle-factor Cox regression and lasso regression were utilized to ensure that the model was not overfitting (Figure 3A and B). Multivariate Cox regression analysis showed that the risk scoring model composed of CCNE1, EZH2, CTHRC1, CBX2, and COL15A1 and miR-9-5p had the smallest Akaike information criterion (AIC) (Figure 3C). A risk survival curve revealed that the samples with a low hazard had better survival status (Figure 3D). The AUCs of the ROC curves in the model were 0.776, 0.745 and 0.789 for 1‐year, 3‐year, and 5‐year OS, respectively (Figure 3E).Figure 3:Construction of Cox regression model and analysis of its diagnostic value. (A) Ten‐time cross‐validation figure. (B) Lasso regression coefficient profiles of 16 genes. (C) Forest plot corresponding to multivariate Cox regression model. (D) Kaplan-Meier risk survival curve of HCC patients. (E) ROC curve assessed the diagnostic efficacy of the risk scoring model.Kaplan-Meier survival analysis suggested that lncRNA FOXD2-AS1, miRNA miR-9-5p and mRNAs STMN1, COL15A1, CCNE2, CTHRC1, LRRC1 and CCNE1 were associated with the prognosis of HCC (Figure 4a–h). To accurately identify the targeted mRNAs of miR-9-5p, the intersections from starBase (seven target genes), TargetScan (1292 target genes), miRDB (1236 target genes) and PicTar (642 target genes) were predicted (Figure 4I), which showed that COL15A1, CCNE2, and STMN1 targeted miR-9-5p in all four databases (Figure 4J). In summary, the FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 regulatory axis was established as a candidate ceRNA array related to the diagnosis and prognosis of HCC.Figure 4:Kaplan-Meier survival analysis and identification of the ceRNA array. (A–H) Kaplan-Meier survival curves of eight genes in HCC patients. (I) Venn diagram of targeted genes of hsa-miR-9-5p. (J) Screening of the ceRNA array.Verification of the candidate ceRNA array as the HCC biomarkerThe FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 array selected from the TCGA database was verified via qRT-PCR. The qRT-PCR results showed that compared with adjacent tissues, FOXD2-AS1, miR-9-5p, STMN1, COL15A1 and CCNE2 were highly expressed in cancer tissues (p<0.05) (Figure 5A). The dual luciferase reporter assay indicated that there were targeted binding relationships between FOXD2-AS1 and miR-9-5p, miR-9-5p and STMN1, but not between miR-9-5p and COL15A1, miR-9-5p or CCNE2 (Figure 5B). The ROC curve demonstrated the diagnostic efficiency of the candidate ceRNA array (FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2), the ceRNA combination (FOXD2-AS1/miR-9-5p/STMN1), and each member (Figure 5C). The AUC value, sensitivity, and specificity of the ceRNA array were 0.926, 84.4%, and 93.1%, respectively, which were all higher than those of the ceRNA combination, which were 0.818, 70.1%, and 79.3% (see Table 5). The ceRNA array formed by the combination of FOXD2-AS1/miR-9-5p/STMN1 combined with COL15A1/CCNE2 could be a diagnostic and prognostic marker for HCC.Figure 5:Verification of FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 array as a HCC marker. (A) Scatter plot corresponding to the differential expression of FOXD2-AS1, miR-9-5p, STMN1, COL15A1 and CCNE2 in HCC tissues and normal tissues. (B) Dual luciferase reporter assay results between miR-9-5p and FOXD2-AS1, miR-9-5p and STMN1, miR-9-5p and COL15A1, miR-9-5p and CCNE2. (C) ROC curves of FOXD2-AS1, miR-9-5p, STMN1, COL15A1, CCNE2, ceRNA (FOXD2-AS1/miR-9-5p/STMN1) and ceRNA array (combination). One-way ANOVA was used for comparison between the two groups. *p<0.05 vs. the control group.Table 5:The results of ROC curve in HCC tissues.GroupAUCStd.p-Value95% CICut-offSensibility/%Specificity/%FOXD2-AS10.6770.0530.0050.573–0.78110.23762.1%66.2%Has-miR-9-5p0.6330.0540.0250.527–0.7399.61662.0%64.7%STMN10.7530.055≤0.0010.646–0.8609.44265.5%80.5%COL15A10.8180.045≤0.0010.730–0.9058.93165.5%83.1%CCNE20.8350.046≤0.0010.744–0.9269.33675.9%83.1%ceRNA0.8180.044≤0.0010.732–0.9050.75170.1%79.3%ceRNA array0.9260.030≤0.0010.867–0.9850.76884.4%93.1%DiscussionHCC is characterized by hidden early clinical symptoms, rapid progression and a poor prognosis, which seriously threatens the lives of cancer patients worldwide [15]. Thus, biomarkers for early diagnosis and prognosis of HCC have become a research hotspot. Numerous ceRNAs have been identified as HCC biomarkers. For example, Dan Cao et al. downloaded HCC data from the TCGA database and constructed a ceRNA network that included lncRNA, mRNA, and miRNA, and used this model to conduct new explorations of molecular mechanisms and prognosis [16]. In recent years, the use of bioinformatics databases and ceRNA theory to explore molecular biomarkers of cancer has become a promising new development. In our study, expression of key genes of a candidate ceRNA array of 96 cancer and 51 adjacent tissues, showed that FOXD2-AS1, miR-9-5p, STMN1, COL15A1, and CCNE2 were upregulated in cancer tissues. The candidate ceRNA array (FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2), has not been reported in previously constructed HCC-related ceRNA networks.FOXD2-AS1, miR-9-5p, STMN1, COL15A1, and CCNE2 have been reported as possible molecular tumor biomarkers. The transcription length of the lncRNA FOXD2-AS1 is 2527 nucleotides, and its abnormal expression in many cancers might contribute to tumor progression through carcinogenesis, cancer cell proliferation, migration and invasion [17]. FOXD2-AS1 plays an oncogenic role in HCC and, can epigenetically silence CDKN1B by recruiting EZH2 to the CDKN1B promoter region [18], and through a ceRNA mechanism with miR-206 and annexin A2 (ANXA2) [19]. FOXD2-AS1 enhanced radiosensitivity in gastric cancer by upregulating SETD1A expression with miR-1913 sponges [20]. MiR-9-5p was found to be highly expressed in HCC and affected the transmission of upstream and downstream molecular regulatory chains as an oncogene. For example, LINC00467 mediated the occurrence of HCC through a ceRNA mechanism with miR-9-5p and peroxisome proliferator-activated receptor alpha (PPARA) [21]. MiR-9-5p was shown to promote the proliferation, invasion and migration of HCC cells by targeting estrogen receptor 1(ESR1) [22]. The miR-9-5p/FOXO1/cytoplasmic polyadenylate binding protein 3(CPEB3) feedforward loop promoted the proliferation and progression of HCC [23]. STMN1, a microtubule-binding protein, regulated the proliferation and migration of HCC cells in vitro and regulated tumor growth in vivo [24]. COL15A1(collagen XV) was highly expressed in tumor tissue of HCC and showed regularity [25]. CCNA2 functions as an oncogene by regulating cell cyclin-dependent kinase (CDK) mediated cell proliferation [26]. These studies indicate that each gene in the ceRNA array model we constructed, bears an important role in tumor progression, and their upregulated expression is consistent with our findings. The scope of targeted ncRNAs was narrowed and the accuracy improved based on ceRNA network relationships to provide specific candidate molecular biomarkers for assessing HCC prognosis.This study found that there were targeted regulatory relationships between FOXD2-AS1 and miR-9-5p, miR-9-5p and STMN1. A report by Bi et al. showed that STMN1 is the target gene of miR-9-5p [27], The findings in this study were in accordance with this report. However, there were no targeted regulatory relationships among miR-9-5p, COL15A1 and CCNE2. Instead, this study found that the ceRNA combination (FOXD2-AS1/miR-9-5p/STMN1) had higher diagnostic ability for HCC than the individual components. Given that the AUC values of COL15A1 and CCNE2 were higher than or equal to the ceRNA combination (FOXD2-AS1/miR-9-5p/STMN1), the candidate ceRNA array with COL15A1 and CCNE2 represented the highest diagnostic value. We speculated that COL15A1 and CCNE2, selected from the ceRNA network, might have some indirect or unverified regulatory relationship with miR-9-5p in addition to the binding relationship of the 3′-UTR region. In view of this, continued exploration of the ceRNA relationships must not only focus on finding direct pathways for lncRNA-miRNA-mRNA interactions, but also focus on indirect ceRNA-distributed network genes.RecommendationsOur research was carried out using HCC tissue. In general, biological indicators in the blood are more useful; therefore, it is suggested that more markers related to ceRNA should be explored and searched in the blood of HCC patients.LimitationsNormally, the regulatory network of ceRNAs in cells tends to be relatively balanced, and if perturbed, this balance will be upset. However, ceRNAs form a complicated network structure composed of numerous genes including lncRNA, circRNA, pseudogenes, miRNA, and mRNA. Moreover, the role of ceRNA in vivo is also constrained by a number of factors, including subcellular location and the abundance of ceRNA components, the interaction with RNA-binding proteins, and the affinity of RNA editing and ceRNA in an endogenous cell environment. Therefore, the uncovered ceRNA mechanism often found can only represent a certain time point of the cell state. As a biomarker, ceRNAs still require substantial research.ConclusionsIn summary, the candidate ceRNA array formed by FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 could be a biomarker for early diagnosis and prognosis of HCC. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png ONCOLOGIE de Gruyter

Early diagnosis and prognosis of hepatocellular carcinoma based on a ceRNA array

Wang, Li-xin; Kong, Ao-ran; Dong, Hui
ONCOLOGIE , Volume 25 (3): 11 – May 1, 2023
Download PDF
11 pages

Loading next page...
 
/lp/de-gruyter/early-diagnosis-and-prognosis-of-hepatocellular-carcinoma-based-on-a-tkgXZO0wQ0
Publisher
de Gruyter
Copyright
© 2023 the author(s), published by De Gruyter, Berlin/Boston
eISSN
1765-2839
DOI
10.1515/oncologie-2023-0088
Publisher site
See Article on Publisher Site

Abstract

IntroductionHepatocellular carcinoma (HCC) has one of the highest incidences and mortality rates worldwide. It involves the carcinogenesis of liver parenchymal cells. In 2020, global cancer data showed 910,000 new liver cancer cases and 830,000 deaths, both of which were in the top 10 of all cancers [1]. Many factors such as infection (HBV, HCV, and parasites), eating food containing aflatoxin, and heavy drinking are related to the occurrence of liver cancer. Despite the development of effective anti-viral drugs, the incidence of HCC is increasing worldwide [2]. Although surgical resection, hepatic artery embolization, radiofrequency ablation, radiotherapy, and chemotherapy have significantly increased the overall survival (OS) of patients with HCC, the prognosis remains very poor, due to a lack of early diagnoses [3, 4]. Therefore, early diagnostic markers and therapeutic targets for HCC need to be urgently explored.Competing endogenous RNA (ceRNA) networks combine the functions of protein coding and non-coding RNAs (ncRNAs), and their mutual communication and co-regulation have important roles in human development and disease processes [5, 6]. The Cancer Genome Atlas (TCGA), Gene Expression Omnibus (GEO), and other public databases are indispensable for bioinformatics analysis, from which researchers can download tumor gene expression profiles for weighted gene co-expression analysis to construct ceRNA networks, which provide important information for tumor marker screening and pathogenic mechanisms [7]. For example, a study found that a ceRNA network composed of three lncRNAs (LINC01106, FOXD2-AS1, and AC103702.2) and three mRNAs (CCDC34, ORC6, and Sox4) was strongly associated with the prognosis of gastric cancer [8]. Three lncRNAs (Dirc3, DNAH10OS, and GK-IT1) were screened from the constructed ceRNA network and were used for evaluation of esophageal cancer risk models [9].Based on HCC information from the TCGA database, we established a lncRNA-miRNA-mRNA ceRNA network to explore the key genes related to the diagnosis and prognosis of HCC, and verified their expression levels and the targeted binding relationship between them in clinical tissues.Materials and methodsAccess to ceRNA network genesThe gene expression profiling information of HCC was obtained from the TCGA database, and differential gene screening was completed using the DESEQ2 package in R4.0.2 software [false discovery rate, (FDR)<0.05; | Log(fold change) |>1], mainly targeting middle lncRNA, miRNA, and mRNA.Construction of the ceRNA networkInteractions between lncRNA, miRNA, and mRNA were analyzed using miRcode [10] and starbase [11] databases (hypergeometric testing, p<0.05; correlation testing, p<0.05). Cytoscape 3.7.2 was used to establish the ceRNA network relationship.The DAVID database was used to perform Gene Ontology and KEGG analyses (p<0.01).Identification of targeted genes of the key miRNATo obtain more reliable targets of core miRNA, TargetScan [12], miRDB [13], and PicTar [14] were searched to identify the intersections with starBase.Validation by qRT-PCR in clinical tissuesBased on the results of bioinformatics analysis, one lncRNA (FOXD2-AS1), one miRNA (has-miR-9-5p), and three mRNAs (STMN1, COL15A1, CCNE2) were identified as candidate genes. Clinical tissue chips were used for gene expression analysis [chip lot numbers: cDNA-HLivH30PG02 and cDNA-HLivH087Su02 (Shanghai Outdo Biotech Co., Ltd. Shanghai, China)], containing a total of 96 cancer tissue samples and 51 normal adjacent tissue samples. The tailing reaction was used for reverse-transcription of miRNA, and the primers were universal. The primers are shown in Table 1. ACTB, encoding β-actin was used for internal reference of lncRNA and mRNA, and U6 was applied for miRNA. The experiment was conducted strictly according to the instructions of the PerfectStart™ Green qPCR SuperMix (TransGen Biotech, Beijing, China) and miRcute Plus miRNA qPCR detection kit (SYBR Green) (Tiangen, Beijing, China). Primer synthesis was completed by Sangon Biotech Co., Ltd., and the primer for miR-9-5p was provided by Tiangen Co., Ltd.Table 1:Primers for lncRNA and mRNA.GenePrimer sequence (5′–3′)FOXD2-AS1FGGCACAGCGGAACGATTGRTGTACCGGCTGAGGCTAGSTMN1FTGGTAGAAATACCCAACGCRGAGGCATCCAAACAAAGCAGTCOL15A1FTGCGACTGAAACAAACAGGTTCARGAAGAAAGCACCCGGAAAAGATGCCNE2FGCATTATGACACCACCGAAGARTAGGGCAATCAATCAATCACAGCF, forward; R, reverse.Tissue cDNA samplesAll HCC tissuecDNA was provided by commercial cDNA microarrays, which were purchased from Shanghai Xinchao Biotech Co., Ltd. The batch numbers of the chips were CDNA-HLIVH30PG02 and cDNA-HLivH087Su02, which contained 96 cancer tissue samples and 51 normal adjacent tissue samples.Dual luciferase reporter assayA wild-type 3′-UTR and its mutants were designed for FOXD2-AS1, STMN1, COL15A1, and CCNE2 binding site sequences to miR-9-5p. The 3′-UTR and its mutants were cloned downstream of the luciferase fragment of a pMiR reporter vector, and then the constructed vector was sequenced. The vector and chemically synthesized miR-9-5p fragment (NC mimics as a control group) were co-transfected into 293T cells. After 48 h, luciferase activity was detected by using a dual-luciferase system (Promega, USA). There were two miR-9-5p binding sites in the 3′-UTR sequence of FOXD2-AS1, and therefore, the base fragments of these two binding sites were connected in series to develop the 3′-UTR and mutant of FOXD2-AS1, which was cloned into a pMiR reporter vector for analysis.Statistical analysisThe GDCRNATools, GGPLOT2, DESEQ2, survival, glmnet, and survminer modules in R4.0.2 software were used to analyze HCC expression profile data from TCGA. Single-factor Cox regression and lasso regression analyses were employed to ensure that the model was not overfitting and all significant genes were integrated into the Cox regression model. Risk survival and the ROC curves were used to evaluate the diagnostic value of the model. Kaplan-Meier survival analysis was applied in batches for each gene in the ceRNA network to assess its hazard. The relative expression of key genes in HCC tissues was calculated via qRT-PCR with the 2⁻△△Ct method, and the difference in the relative expression between the HCC cancer group and the para-cancer group was analyzed using an independent sample t-test. The ROC curve was drawn according to qRT-PCR detection of the tissue cDNA chip. The diagnostic value of all important members, ceRNA combinations and candidate ceRNA arrays was evaluated with the AUC value. The statistical analysis and plotting were performed using IBM SPSS Statistics 25.0. and GraphPad Prism 8.ResultsLncRNA-miRNA-mRNA ceRNA in HCCIn this study, HCC transcriptome data from 424 samples (adjacent: 50 cases, cancer: 374 cases), miRNA data from 425 samples (adjacent: 50 cases, cancer: 375 cases), and related clinical information were downloaded from the TCGA database. Survival analysis and univariate analysis were performed on the clinical information of the 366 HCC patients in the TCGA database, which revealed that T stage, M stage and N stage were associated with survival of HCC patients (see Table 2).Table 2:Clinical correlation analysis of TCGA database baseline data.GroupNCumulative survival rateLog-rankUnivariate analysis3 years5 yearsX2p-ValueHR95% CIp-ValueAge, years3661.3840.239≤6017562.1%54.6%1＞6019162.3%47.7%1.2350.868–1.7580.240Gender3661.2040.273Female11858.1%45.3%1Male24865.1%52.9%0.8180.571–1.1720.273Grade3661.2750.735G15567.8%55.2%1G217862.1%50.1%1.1830.699–2.0030.531G312159.2%48.1%1.2380.713–2.1490.448G4 TNM12––1.7120.632–4.6360.290T36639.373≤0.001T117977.1%64.6%1T29559.4%43.2%1.5200.959–2.4100.075T37944.4%33.4%2.7151.775–4.151≤0.001T413005.5052.742–11.053≤0.001N3664.6800.096N025066.4%58.8%1N1433.3%-2.1110.517–8.6270.298NX11253.8%27.4%1.4571.002–2.1190.049M36610.8100.004M026467.4%58.8%1M14004.2631.340–13.5660.014MX9852,6%31.9%1.5471.058–2.2610.024Stage36634.761≤0.001Ⅰ17477.4%65.8%1Ⅱ9557.6%41.9%1.6601.045–2.6390.032Ⅲ8842.0%31.6%2.9221.919–4.450≤0.001Ⅳ914.3%04.6842.097–10.460≤0.001A total of 136 DE-lncRNAs (104 upregulated, 32 downregulated) (Figure 1A), 128 DE-miRNAs (107 upregulated, 21 downregulated) (Figure 1B), and 2028 DE-miRNAs (1222 upregulated, 806 downregulated) (Figure 1C) were obtained using gene differential expression analysis.Figure 1:Competing endogenous RNA-related gene screening. (A) Heat map and volcano plot of differentially expressed lncRNAs. (B) Heat map and volcano plot of differentially expressed miRNAs. (C) Heat map and volcano plot of differentially expressed mRNAs.The differentially expressed lncRNAs (n=8), miRNAs (n=5), and prognostic mRNAs (n=21) were used to construct a ceRNA network (Figure 2A) (see Table 3).Figure 2:CeRNA network and functional enrichment analysis. (A) Competing endogenous RNA network in HCC. (B) Results for KEGG enrichment analysis of the mRNAs in the ceRNA network.Table 3:LncRNA, miRNA and mRNA that constituted the ceRNA network.RNA typeGene IDLncRNAAC092171.2, AC239868.2, CRNDE, PVT1, FOXD2-AS1, AL445524.1, LINC00511, AC111000.4miRNAhsa-miR-96-5p, hsa-miR-182-5p, hsa-miR-424-5p, hsa-miR-217, hsa-miR-9-5pmRNAANLN, KIF23, CCNE1, RBM24, GNAL, SLC6A9, ZIC2, STMN1, CTHRC1, CBX2, DTL, EZH2, CCNE2, APLN, SNCG, GPC3, LRRC1, NRCAM, IGF2BP1, TSPAN5, COL15A1The detailed regulatory relationships of lncRNA-miRNA and miRNA-mRNA are described in Table 4. Functional enrichment analysis of GO and KEGG showed that 21 mRNAs were mainly enriched in 14 signaling pathways, of which “microRNAs in cancer pathway” was the most important (Figure 2B).Table 4:Regulatory relationships between genes in ceRNA network.lncRNAProtein-coding RNAmiRNAHypergeometric test p-ValueCorrelation p-ValueAC092171.2ANLNhsa-miR-96-5p0.0002331.85E-30hsa-miR-424-5phsa-miR-182-5pKIF23hsa-miR-2170.0064132.57E-28hsa-miR-424-5CCNE1hsa-miR-424-5p0.0011791.37E-24RBM24hsa-mi R-424-5p0.003890.000133GNALhsa-miR-424-5p0.0057453.40E-06SLC6A9hsa-miR-96-5p0.0001091.20E-09hsa-miR-182-5pAC239868.2ZIC2hsa-miR-96-5p0.0041161.15E-17STMN1hsa-miR-9-5p0.0074184.86E-21CTHRC1hsa-miR-9-5p0.0005890.0000108CBX2hsa-miR-96-5p0.0001261.97E-12CRNDEDTLhsa-miR-2170.0063071.69E-09EZH2hsa-miR-2170.0003651.58E-11CCNE2hsa-miR-9-5p0.0033760.000215PVT1APLNhsa-miR-424-5p0.0004450.00073KIF23hsa-miR-424-5p0.0001091.24E-06CTHRC1hsa-miR-9-5p0.001371.23E-13SNCGhsa-miR-424-5p0.0004891.41E-19GNALhsa-miR-424-5p0.0036841.15E-08FOXD2-AS1STMN1hsa-miR-9-5p0.0019441.01E-26CCNE1hsa-miR-9-5p0.0077148.93E-29AL445524.1GPC3hsa-miR-96-5p0.0011591.23E-11hsa-miR-182-5pANLNhsa-miR-96-5p0.0011910.000745hsa-miR-182-5phsa-miR-217LRRC1hsa-miR-9-5p0.0003050.000682NRCAMhsa-miR-182-5p0.0042920.0003LINC00511IGF2BP1hsa-miR-9-5p0.0003939.07E-17CTHRC1hsa-miR-9-5p0.0001681.84E-19CBX2hsa-miR-424-5p0.000149.90E-16SNCGhsa-miR-424-5p0.0017472.04E-08GNALhsa-miR-424-5p0.0017471.27E-23TSPAN5hsa-miR-424-5p0.0008830.004075AC111000.4COL15A1hsa-miR-9-5p0.0009095.40E-05Determination of the ceRNA arraySingle-factor Cox regression and lasso regression were utilized to ensure that the model was not overfitting (Figure 3A and B). Multivariate Cox regression analysis showed that the risk scoring model composed of CCNE1, EZH2, CTHRC1, CBX2, and COL15A1 and miR-9-5p had the smallest Akaike information criterion (AIC) (Figure 3C). A risk survival curve revealed that the samples with a low hazard had better survival status (Figure 3D). The AUCs of the ROC curves in the model were 0.776, 0.745 and 0.789 for 1‐year, 3‐year, and 5‐year OS, respectively (Figure 3E).Figure 3:Construction of Cox regression model and analysis of its diagnostic value. (A) Ten‐time cross‐validation figure. (B) Lasso regression coefficient profiles of 16 genes. (C) Forest plot corresponding to multivariate Cox regression model. (D) Kaplan-Meier risk survival curve of HCC patients. (E) ROC curve assessed the diagnostic efficacy of the risk scoring model.Kaplan-Meier survival analysis suggested that lncRNA FOXD2-AS1, miRNA miR-9-5p and mRNAs STMN1, COL15A1, CCNE2, CTHRC1, LRRC1 and CCNE1 were associated with the prognosis of HCC (Figure 4a–h). To accurately identify the targeted mRNAs of miR-9-5p, the intersections from starBase (seven target genes), TargetScan (1292 target genes), miRDB (1236 target genes) and PicTar (642 target genes) were predicted (Figure 4I), which showed that COL15A1, CCNE2, and STMN1 targeted miR-9-5p in all four databases (Figure 4J). In summary, the FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 regulatory axis was established as a candidate ceRNA array related to the diagnosis and prognosis of HCC.Figure 4:Kaplan-Meier survival analysis and identification of the ceRNA array. (A–H) Kaplan-Meier survival curves of eight genes in HCC patients. (I) Venn diagram of targeted genes of hsa-miR-9-5p. (J) Screening of the ceRNA array.Verification of the candidate ceRNA array as the HCC biomarkerThe FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 array selected from the TCGA database was verified via qRT-PCR. The qRT-PCR results showed that compared with adjacent tissues, FOXD2-AS1, miR-9-5p, STMN1, COL15A1 and CCNE2 were highly expressed in cancer tissues (p<0.05) (Figure 5A). The dual luciferase reporter assay indicated that there were targeted binding relationships between FOXD2-AS1 and miR-9-5p, miR-9-5p and STMN1, but not between miR-9-5p and COL15A1, miR-9-5p or CCNE2 (Figure 5B). The ROC curve demonstrated the diagnostic efficiency of the candidate ceRNA array (FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2), the ceRNA combination (FOXD2-AS1/miR-9-5p/STMN1), and each member (Figure 5C). The AUC value, sensitivity, and specificity of the ceRNA array were 0.926, 84.4%, and 93.1%, respectively, which were all higher than those of the ceRNA combination, which were 0.818, 70.1%, and 79.3% (see Table 5). The ceRNA array formed by the combination of FOXD2-AS1/miR-9-5p/STMN1 combined with COL15A1/CCNE2 could be a diagnostic and prognostic marker for HCC.Figure 5:Verification of FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 array as a HCC marker. (A) Scatter plot corresponding to the differential expression of FOXD2-AS1, miR-9-5p, STMN1, COL15A1 and CCNE2 in HCC tissues and normal tissues. (B) Dual luciferase reporter assay results between miR-9-5p and FOXD2-AS1, miR-9-5p and STMN1, miR-9-5p and COL15A1, miR-9-5p and CCNE2. (C) ROC curves of FOXD2-AS1, miR-9-5p, STMN1, COL15A1, CCNE2, ceRNA (FOXD2-AS1/miR-9-5p/STMN1) and ceRNA array (combination). One-way ANOVA was used for comparison between the two groups. *p<0.05 vs. the control group.Table 5:The results of ROC curve in HCC tissues.GroupAUCStd.p-Value95% CICut-offSensibility/%Specificity/%FOXD2-AS10.6770.0530.0050.573–0.78110.23762.1%66.2%Has-miR-9-5p0.6330.0540.0250.527–0.7399.61662.0%64.7%STMN10.7530.055≤0.0010.646–0.8609.44265.5%80.5%COL15A10.8180.045≤0.0010.730–0.9058.93165.5%83.1%CCNE20.8350.046≤0.0010.744–0.9269.33675.9%83.1%ceRNA0.8180.044≤0.0010.732–0.9050.75170.1%79.3%ceRNA array0.9260.030≤0.0010.867–0.9850.76884.4%93.1%DiscussionHCC is characterized by hidden early clinical symptoms, rapid progression and a poor prognosis, which seriously threatens the lives of cancer patients worldwide [15]. Thus, biomarkers for early diagnosis and prognosis of HCC have become a research hotspot. Numerous ceRNAs have been identified as HCC biomarkers. For example, Dan Cao et al. downloaded HCC data from the TCGA database and constructed a ceRNA network that included lncRNA, mRNA, and miRNA, and used this model to conduct new explorations of molecular mechanisms and prognosis [16]. In recent years, the use of bioinformatics databases and ceRNA theory to explore molecular biomarkers of cancer has become a promising new development. In our study, expression of key genes of a candidate ceRNA array of 96 cancer and 51 adjacent tissues, showed that FOXD2-AS1, miR-9-5p, STMN1, COL15A1, and CCNE2 were upregulated in cancer tissues. The candidate ceRNA array (FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2), has not been reported in previously constructed HCC-related ceRNA networks.FOXD2-AS1, miR-9-5p, STMN1, COL15A1, and CCNE2 have been reported as possible molecular tumor biomarkers. The transcription length of the lncRNA FOXD2-AS1 is 2527 nucleotides, and its abnormal expression in many cancers might contribute to tumor progression through carcinogenesis, cancer cell proliferation, migration and invasion [17]. FOXD2-AS1 plays an oncogenic role in HCC and, can epigenetically silence CDKN1B by recruiting EZH2 to the CDKN1B promoter region [18], and through a ceRNA mechanism with miR-206 and annexin A2 (ANXA2) [19]. FOXD2-AS1 enhanced radiosensitivity in gastric cancer by upregulating SETD1A expression with miR-1913 sponges [20]. MiR-9-5p was found to be highly expressed in HCC and affected the transmission of upstream and downstream molecular regulatory chains as an oncogene. For example, LINC00467 mediated the occurrence of HCC through a ceRNA mechanism with miR-9-5p and peroxisome proliferator-activated receptor alpha (PPARA) [21]. MiR-9-5p was shown to promote the proliferation, invasion and migration of HCC cells by targeting estrogen receptor 1(ESR1) [22]. The miR-9-5p/FOXO1/cytoplasmic polyadenylate binding protein 3(CPEB3) feedforward loop promoted the proliferation and progression of HCC [23]. STMN1, a microtubule-binding protein, regulated the proliferation and migration of HCC cells in vitro and regulated tumor growth in vivo [24]. COL15A1(collagen XV) was highly expressed in tumor tissue of HCC and showed regularity [25]. CCNA2 functions as an oncogene by regulating cell cyclin-dependent kinase (CDK) mediated cell proliferation [26]. These studies indicate that each gene in the ceRNA array model we constructed, bears an important role in tumor progression, and their upregulated expression is consistent with our findings. The scope of targeted ncRNAs was narrowed and the accuracy improved based on ceRNA network relationships to provide specific candidate molecular biomarkers for assessing HCC prognosis.This study found that there were targeted regulatory relationships between FOXD2-AS1 and miR-9-5p, miR-9-5p and STMN1. A report by Bi et al. showed that STMN1 is the target gene of miR-9-5p [27], The findings in this study were in accordance with this report. However, there were no targeted regulatory relationships among miR-9-5p, COL15A1 and CCNE2. Instead, this study found that the ceRNA combination (FOXD2-AS1/miR-9-5p/STMN1) had higher diagnostic ability for HCC than the individual components. Given that the AUC values of COL15A1 and CCNE2 were higher than or equal to the ceRNA combination (FOXD2-AS1/miR-9-5p/STMN1), the candidate ceRNA array with COL15A1 and CCNE2 represented the highest diagnostic value. We speculated that COL15A1 and CCNE2, selected from the ceRNA network, might have some indirect or unverified regulatory relationship with miR-9-5p in addition to the binding relationship of the 3′-UTR region. In view of this, continued exploration of the ceRNA relationships must not only focus on finding direct pathways for lncRNA-miRNA-mRNA interactions, but also focus on indirect ceRNA-distributed network genes.RecommendationsOur research was carried out using HCC tissue. In general, biological indicators in the blood are more useful; therefore, it is suggested that more markers related to ceRNA should be explored and searched in the blood of HCC patients.LimitationsNormally, the regulatory network of ceRNAs in cells tends to be relatively balanced, and if perturbed, this balance will be upset. However, ceRNAs form a complicated network structure composed of numerous genes including lncRNA, circRNA, pseudogenes, miRNA, and mRNA. Moreover, the role of ceRNA in vivo is also constrained by a number of factors, including subcellular location and the abundance of ceRNA components, the interaction with RNA-binding proteins, and the affinity of RNA editing and ceRNA in an endogenous cell environment. Therefore, the uncovered ceRNA mechanism often found can only represent a certain time point of the cell state. As a biomarker, ceRNAs still require substantial research.ConclusionsIn summary, the candidate ceRNA array formed by FOXD2-AS1/miR-9-5p/STMN1/COL15A1/CCNE2 could be a biomarker for early diagnosis and prognosis of HCC.

Journal

ONCOLOGIE – de Gruyter

Published: May 1, 2023

Keywords: biomarker; competitive endogenous RNA array; FOXD2-AS1/miR-9-5p/ STMN1/COL15A1/CCNE2; hepatocellular carcinoma

There are no references for this article.