Tissue-based MicroRNA Biomarkers In Three Histological Subtypes Of Lung Cancer:Discovery, Validation And Target Screening

Part I Discovery of tissue-based microRNA biomarkers in three histological subtypes of lung cancerObjective:Discover tissue-based microRNA biomarkers in discriminating three histological subtypes of lung cancerMethods:Laser-capture microdissection (LCM) was applied to isolate pure targeted cells on each of126frozen surgical specimens from44normal tissues,36adenocarcinoma (AC),30squamous carcinoma (SQ) and16small cell lung cancer (SCLC) tissues. Agilent microRNA microarray platform was employed on the LCM-selected target cells to discover the candidate microRNA biomarkers. After quantile normalization, the microarray data was analyzed with one-way analysis of variance (ANOVA), hierarchical clustering and prediction analysis of microarray (PAM and WEAK). Furthermore,7candidate microRNAs selected by both PAM and WEAK were analyzed with receiver operator characteristic curve (ROC).Results:16differentially expressed microRNAs in the three histological subtypes of lung cancer were discovered by ANOVA analysis. Using PAM and WEAK methods in the prediction analysis of microarray, two classifiers of17and13microRNAs with prediction accuracies of87%-90%were identified in discriminating the three subtypes of lung cancer. The predicted outcomes using the two microRNA classifiers were well consistent with the pathological diagnosis. Of the components in the two classifiers,7microRNAs (hsa-miR-25, hsa-miR-205, hsa-miR-34a, hsa-miR-375, hsa-miR-29a, hsa-miR-29b and hsa-miR-27a) were selected by the two employed classifications. The diagnostic accuracy (AUC value) of the7individual microRNAs in discriminating the three subtypes of lung cancer was between0.623and0.992.Conclusions:The newly discovered microRNA classifiers and the selected7candidate microRNAs could accurately discriminate the three histological subtypes of lung cancer. The results indicate that these microRNAs may involve in the subtypes of lung carcinogenesis. Part Ⅱ Validation of tissue-based microRNA biomarkers in three histological subtypes of lung cancerObjective:Validate the7candidate microRNAs, establish logistic regression models in discriminating the three histological subtypes of lung cancer and analyze their prognostic values.Methods:In207formalin-fixed and paraffin-embedded (FFPE) tissue specimens, quantitative RT-PCR (Taqman) was performed on7candidate microRNAs that could discriminate the three histological subtypes of lung cancer. In the training dataset (30normal tissues,30AC,27SQ and32SCLC), unpaired t-test and ROC curve analysis were performed on7individual microRNAs to determine the significance and diagnostic accuracy. Furthermore, stepwise logistic regression analysis was applied to establish the regression models with the best combination of multiple diagnostic microRNAs. In the validation dataset (22normal tissues,22AC,23SQ and21SCLC), ROC curve analysis were performed on the logit (p) values of the constructed regression models to validate their diagnostic values in discriminating the three subtypes of lung cancer. Additionally, Kaplan-Meier and log-rank tests were used to analyze the overall survival duration in three subtypes of lung cancer, Cox regression analysis was applied for the multivariate analysis to assess the association of7microRNA biomarkers with overall survival (OS).Results:In the training dataset, each of the logistic regression models with the best combinations of7microRNA biomarkers yielded high accuracies (AUC>0.9) in discriminating the three subtypes of lung cancer. In the validation dataset, the regression models were validated with the similar diagnostic accuracies (AUC-0.9). Significant variance and fold change analysis on7individual microRNAs indicate that hsa-miR-205is a top discriminator in squamous carcinoma, while hsa-miR-375is a key player in small cell lung cancer. MST of each type of lung cancer is (month): AC65, SQ53and SCLC18. Log Rank test showed the significant difference on the OS in the comparisons of AC vs. SCLC (P-value=0.001) and SQ vs. SCLC (P-value=0,002), while no significant difference between AC and SQ comparison (P-value=0.715). Furthermore, no significant correlation was observed between individual microRNA and the OS in each of three subtype lung cancers by Cox regression multivariate analysis (all P-values>0.05).Conclusions:The validated regression models with the best combinations of7 microRNA biomarkers could accurately discriminate the three subtypes of lung cancer. The results further imply that the7validated microRNA biomarkers associate with the three subtypes of lung carcinogenesis. Of them, hsa-miR-205and hsa-miR-375are top discriminators for squamous carcinoma and small cell lung cancer, respectively. The overall survival of AC and SQ was significantly better than that of SCLC patients. All7microRNA biomarkers had no effect on OS in all the three subtypes of lung cancer. Part III Prediction of the target genes for7microRNA biomarkers and screening the targets of hsa-miR-205and hsa-miR-375Objective:Predict the target genes for7microRNA biomarkers and screen the targets of hsa-miR-205and hsa-miR-375in the three histological subtypes of lung cancerMethods:The target genes for each of7microRNA biomarkers were predicted through the gateway miRecords. The genes that were predicted by at least4of11databases were selected as targets. The KEGG database was used to map the predicted targets of microRNAs to22lung cancer-associated pathways. In133frozen lung tissues (48normal tissues,34AC,37SQ and14SCLC), quantitative RT-PCR (SYBER-Green) was performed on the target genes of hsa-miR-205and hsa-miR-375. The corrected p value<0.05in the unpaired t-test and fold change≥2.0were used to determine the statistical significance.Results:Putative targets for7microRNA biomarkers were identified through miRecords and KEGG databases. The numbers of their target genes (including the same gene in more than one signal pathways) are:108targets for hsa-miR-205,37for hsa-miR-375,115for hsa-miR-25,209for hsa-miR-29a,216for hsa-miR-29b,254for hsa-miR-27a and150for hsa-miR-34a. On the tissue level,12significant differential expressed target genes of hsa-miR-205were identified. These target genes are NCAM1and SMAD4in AC; ATP2B2, AXIN2, CACNA2D2, CDH3, E2F1, NCAM1, RELN, SMAD4, STAG1and RUNX1in SQ and NCAM1、KIT and CDC27in SCLC. For hsa-miR-375,5significant differential expressed target genes were identified. The genes include ITPKB in AC; CACNG2, ITPKB, PIAS1and RUNX1in SQ and ITPKB, NLK and PIAS1in SCLC.Conclusions:The combination of miRecords and KEGG database is a good method to predict and select signaling-related target genes for microRNA. The expression of the target mRNA and its microRNA should be both tested in the tissue level to screen the target genes. The same microRNA may play different roles in the different subtypes of lung cancer. The target genes with the positive and negative correlation of its microRNA expression shall be selected for further functional studies.