Discovery And Validation Of MiRNA Biomarkers In Colorectal Cancer And Precancer

IntroductionColorectal adenocarcinoma is a major cause of cancer mortality worldwide. Detecting the cancer at early stages has significant therapeutic values. Therefore, there is an unmet need to identify new class of biomarkers of colorectal early lesions and colorectal adenocarcinoma.Ideal tumor markers shall have good sensitivity, specificity and can be easy to detect. MicroRNA (miRNA),18-to 25-nucleotides, noncoding RNA molecules, has met the criteria mentioned above. Demonstrated evidences show that miRNAs involve in the regulation of carcinogenesis and function as both tumour suppressors and oncogenes. Furthermore, miRNAs have major potentials as diagnostic and prognostic biomarkers, as they strongly associate with the clinical outcomes. Additionally, miRNAs have advantages over mRNAs as cancer biomarkers, since they are very stable in vitro and long-lived in vivo.Molecular profiling of clinical tissue specimens is frequently complicated by their cellular heterogeneity. Laser capture microdissection (LCM) has successfully been used to tackle this problem by isolating pure cell populations from tissue sections. However, the quantity and quality of material recovered after LCM is often still limited. Therefore, combination of LCM and whole genome analysis has not been widely applied to discover miRNA biomarkers in solid tumors. So far, the large majority of published miRNA expression studies utilized whole tumor tissues without separating the truly transformed cancerous cells from those other cell types commonly present within a solid tumor. Analysis of such complex tissues could conceal the specific signature of the particular cell type of interest.In this study, we improved tissue preparation for LCM and combined LCM with genome-wide miRNA analysis to discovery the potential miRNA biomarkers at the multisteps of the colorectal carcinogenesis using frozen surgical specimens. We evaluated the miRNA expression profiles of colorectal adenocarcinoma and the precancerous lesions to study their potential role in the tumor formation and molecular classification in colorectal carcinoma. We then validated the candidate miRNA biomarkers in the colorectal adenocarcinoma and the precancerous lesions using FFPE (formalin-fixed, paraffin-embedded) surgical specimens. We finally explored the clinical utilities of the validated miRNA biomarkers for early cancer detection in FFPE biopsy tissues from colonoscopy.PartⅠImprovement of tissue preparation for laser capture microdissection: application for cell type-specific miRNA expression profilingObjective:Improve the tissue preparation for laser capture microdissection in order that this technique can be applied for cell type-specific miRNA expression profiling.Methods:Six different experiments on tissue preparations were performed. Total RNA was extracted by using mirVana miRNA isolation kit. The concentration was quantified by NanoDrop 1000 Spectrophotometer. The quality control of RNA was performed by a 2100 Bioanalyzer using the RNA 6000 Pico LabChip kit. Using the improved protocol for tissue preparation, laser capture microdissection (LCM) was performed on 101 frozen surgical specimens of colorectal tissues (n=43 normal, n=14 adenomas and n= 44 carcinomas) to assess the RNA quality. Paired t-test was performed for significance analysis. Human miRNA microarrays were used to compare the expression profiles of 18 colorectal tissues (n=6 normal, n=6 adenomas and n=6 carcinomas) between LCM selected pure epithelial cells versus stromal cells. GeneSpring GX10 software was applied for quantile normalization. Unpaired t-test with Benjamini-Hochberg correction and one-way analysis of variance (ANOVA) were used for differential miRNA expression analysis. Hierarchical clustering was performed with Pearson correlation using the differentially expressed miRNAs.Results:We found that the ethanol fixation of tissue sections for 2 hours had the maximum improvement of RNA quality (1.8 fold, p=0.0014) and quantity (1.5 fold, p=0.066). Overall, the quality (RNA integrity number, RIN) for the microdissected colorectal tissues was 5.2±1.5 (average±SD) for normal (n=43),5.7±1.1 for adenomas (n=14) and 7.2±1.2 for carcinomas (n=44). We then compared miRNA expression profiles of 18 colorectal tissues (6 normal,6 adenomas and 6 carcinomas) between LCM selected pure epithelial cells versus stromal cells using Agilent miRNA microarrays. We identified 51 differentially expressed miRNAs (p<=0.001) between these two cell types. Additionally, we found that 26 miRNAs in the epithelial cells could differentiate adenomas from normal and carcinomas. However, the miRNAs in the stromal and mixed cells could not separate adenomas from normal tissues.Conclusions:The improvement of the tissue preparation for laser capture microdissection can increase the quality and quantity of RNA, which meet the needs of cell type-specific miRNA expression profiling. Our study demonstrates the feasibility and potential power of discovering cell type-specific miRNA biomarkers in complex tissue using combination of LCM with genome-wide miRNA analysis.PartⅡDiscovery of miRNA biomarkers in the epithelial cells during the transformation and progression of colorectal cancerObjective:Discover miRNA biomarkers in LCM-selected epithelial cells covered every stage of colorectal carcinogenesis.Methods:The optimized LCM protocol was applied to isolate pure epithelial cells from 225 frozen surgical specimens of colorectal normal, precancerous lesion and invasive carcinoma tissues. Genome-wide miRNA analysis was performed to determine the expression profiles in these LCM-selected epithelial cells. GeneSpring GX10 software was used for quantile normalization. Unpaired t-test with Benjamini-Hochberg correction and one-way analysis of variance (ANOVA) were applied for the differential expression analysis. Hierarchical clustering was performed with Pearson correlation using the differentially expressed miRNAs. Three prediction supervised classification algorithms (prediction analysis of microarray, genetic algorithm-SVM and one-loop Naive Bayesian) were employed to analyze the data acquired on the microarrays. Subsequently, quantitative RT-PCR was used to verfy the miRNA expression profiles.Results:42 miRNA signatures were discovered in the transformation and progression of colorectal cancer. Two classifiers of miRNA signatures were identified for predicting colorectal cancer. A minimal set of 20 miRNAs could distinct non-neoplasm polyps and neoplasm polyps with the highest accuracy of 96.9%, while another minimal set of 14 miRNAs could discriminate benign neoplasm from malignant neoplasm with the highest accuracy of 99.9%. The average quantitative correlation (R) of fold change between Agilent miRNA microarrays and quantitative RT-PCR was 0.980.Conclusions:The candidate miRNA biomarkers were discovered in the transformation and progression of colorectal cancer. Such biomarkers could accurately discriminate non-neoplasm polyps from neoplasm polyps and benign neoplasm from malignant neoplasm. The predicted outcomes using the miRNA classifiers were well consistent with the pathological diagnosis.PartⅢValidation of the candidate miRNA biomarkers in discriminating colorectal benign neoplasm from malignant neoplasmObjective:Validate the candidate miRNA biomarkers in discriminating colorectal benign neoplasm from malignant neoplasm.Methods:Quantitative RT-PCR was performed on 14 candidate miRNA biomarkers that could discriminate benign neoplasm from malignant neoplasm using 185 FFPE colorectal tissue specimens, including 124 surgical specimens and 61 biopsy sepcimens. Unpaired t-test was performed for significance analysis. Receiver operating characteristic (ROC) curve analysis was performed to determine the specificity and sensitivity of individual miRNA as diagnostic biomarkers. Stepwise logistic regression analysis was performed to determine the specificity and sensitivity of combined miRNAs as diagnostic biomarkers.Results:The early colorectal carcinomas could be distinguished from benign tumors by the validated miRNA biomarkers. Of the 14 candidate miRNA biomarkers,7 (hsa-miR-125b, hsa-miR-7, hsa-miR-218, hsa-miR-375, hsa-miR-424, hsa-miR-92a, hsa-miR-99a) were validated for distincting benign neoplasm from malignant neoplasm in the FFPE surgical specimens. The combination of hsa-miR-92a and hsa-miR-375 yielded the sensitivity of 95% and spesitivity of 88% in discriminating high-grade intraepithelial neoplasm from Dukes'A carcinoma. Furthermore, these miRNAs have the ability to differentiate the high-grade intraepithelial neoplasm from carcinomas in the biopsy tissues from colonoscopy with sensitivity of 93% by hsa-miR-92a and spesificity of 94% by hsa-miR-7.Conclusions:The early colorectal carcinoma could be accurately discriminated from benign tumor by 7 validated miRNA biomarkers. Colorectal carcinomas could be distincted from the high-grade intraepithelial neoplasm in the colonoscopy biopsy tissues using these validatd miRNA biomarkers.