Comprehensive workflow for the identification and validation of micrornas using high-throughput qPCR technologies: the colorectal cancer experience

Introduction MicroRNAs (miRNAs) are small non-coding RNAs involved in the development of various cancers. qPCR low density arrays, allowing to simultaneously evaluate hundreds of miRNAs, are commonly used to investigate their expression in the discovery phase. In order to prevent false-positive results, different factors should be taken into consideration in the workflow involved in the identification and validation of miRNAs as cancer biomarkers. Data normalization is one of the most crucial pre-processing steps for high-throughput qPCR analysis [1]. In addition, among the pre-analytical factors that can influence the miRNAs expression, the haemolysis (rupture of erythrocytes) is one of the most relevant. Aim To implement a comprehensive workflow for the identification and the subsequent validation of miRNAs as biomarkers for the early detection of colorectal cancer by: (i) defining a proper data-normalization strategy for high-throughput qPCR data and (ii) assessing the influence of haemolysis on miRNAs expression. Methods Plasma samples from 60 subjects positive to the fecal occult blood test (FIT+) were processed with TaqMan® Array MicroRNA 384-well Card A. Starting from this data we computed the relative expression of each miRNA according to the mean expression value of all the miRNAs detected in all samples [3]. The best subset of miRNAs that resemble this value was selected as normalization factor to use also in the validation phase [4,5]. The miRNAs relative expression was used as response variable to identify candidate miRNAs associated with precancerous/cancerous lesions. Moreover, we set-up an in-vitro controlled haemolysis experiment to investigate if miRNAs expression is influenced by different level of haemolysis. Briefly, red blood cells were artificially introduced in a heamolysis-free plasma sample by performing serial dilutions and single qPCR were performed for miRNAs quantification. We computed the 95% Simultaneous Confidence Interval (SCI) [6] of the Relative Quantity (RQ) and miRNAs showing an upper limit of the 95% SCI(RQ) ≥ 2 or a lower limit of the 95% SCI(RQ) ≤ 0.5 were considered statistically influenced by haemolysis. Results We identified a subset of 4 reference miRNAs suitable for data normalization and 7 miRNAs significantly different in subjects with precancerous/cancerous lesions versus subjects without lesions. MiRNAs known as haemolysis-related in literature [2] were confirmed also in our in-vitro controlled experiment whereas our reference and candidate miRNAs seem to be not influenced by haemolysis. Conclusion The developed workflow allowed: (i) the identification of reference miRNAs to be used for data normalization together with the identification of candidate miRNAs and (ii) the evaluation of the haemolsyis influence on the expression of these miRNAs. We are now analyzing these candidate miRNAs on a prospective validation cohort of FIT+ subjects using a qPCR customized assay. Acknowledgements This work was supported by grant from Associazione Italiana per la Ricerca sul Cancro (AIRC) (Grant No. 12162). Bibliografia 1. Deo A, Carlsson J and Lindlof A. J Bioinform Comput Biol 2011; 9:795-812 2. Kirschner MB, Kao SC, Edelman JJ, et al. PlosOne. 2011; 6:e24145 3. Mestdagh P, Van Vlierberghe P, De Weer A, et al. Genome Biol 2009;10:R64 4. Pizzamiglio S, Bottelli S, Ciniselli CM, et al. Int J Cancer 2014; 134:2016-8 5. Verderio P, Bottelli S, Ciniselli CM, et al. Anal Biochem. 2014; 461:7-9 6. Pizzamiglio S, Cossa G, Gatti L, et al. Oncol Rep. 2010;23:853-60