Methods in Molecular Biology DOI 10.1007/7651_2015_247 © Springer Science+Business Media New York 2015
Analysis of microRNA Microarrays in Cardiogenesis Diego Franco, Fernando Bonet, Francisco Hernandez-Torres, Estefania Lozano-Velasco, Francisco J. Esteban, and Amelia E. Aranega Abstract microRNAs are a subclass of noncoding RNAs which have been demonstrated to play pivotal roles in multiple cellular mechanisms. microRNAs are small RNA molecules of 22–24 nt in length capable of modulating protein translation and/or RNA stability by base-priming with complementary sequences of the mRNAs, normally at the 30 untranslated region. To date, over 2,000 microRNAs have been already identified in humans, and orthologous microRNAs have been also identified in distinct animals and plants ranging a wide vast of species. High-throughput analyses by microarrays have become a gold standard to analyze the changes on microRNA expression in normal and pathological cellular or tissue conditions. In this chapter, we provide insights into the usage of this uprising technology in the context of cardiac development and disease. Keywords: microRNA, Microarrays, Cardiac development, Meta-analyses
1
Background Cardiac development is a complex process in which multiple cell types are involved (1–3). From the early stages of cardiogenic specification soon after gastrulation, the heart progressively acquires a tubular shape formed by an inner endothelial lining and an outer myocardium layer (4). The heart is the first organ to display left-right asymmetry (5, 6) and more importantly, it is the first organ to be functional during organogenesis (2). Soon after the heart start to pump, atrial and ventricular chambers are progressively configured and valve primordia are formed in the intertwining areas, i.e., outflow tract, atrioventricular canal, and inflow tract (2). Subsequently, the heart is separated and each of these structures becomes divided into left and right parts (7, 8). Over the last decades, we have gained crucial insights into the molecular mechanisms that govern cardiac morphogenesis. Given the complexity of cardiac morphogenesis, multiple pathways and transcriptional factors are involved during cardiogenesis at different stages, as recently reviewed (9, 10). In the last years, a novel layer of complexity is emerging in the cardiovascular
Diego Franco et al.
development field, namely the post-transcriptional regulatory networks driven by noncoding RNAs. Among noncoding RNAs, microRNAs have emerged as a pivotal mechanism is cardiovascular development, since targeted deletion of the microRNAs processing ribonuclease Dicer resulted in severe cardiovascular defects. Moreover, targeted deletion of single microRNAs, such as miR-1 and miR-126, respectively, also resulted in severe cardiovascular development impairment (11, 12). In view of the important contribution of distinct microRNAs in cardiovascular field, great efforts have been devoted to map the microRNA microarray fingerprints of distinct normal and abnormal cardiovascular contexts (see for a recent review (13–15)). In this chapter, we provide insights into the usage of microRNA microarray analyses in the field of cardiovascular development, as well as insights as how to proceed further beyond the classical microarray approach, such as microarray meta-analyses.
2
Materials and Methods Initial steps to ensure appropriate microRNA microarray analyses start already on the experimental design. Considering a simple experiment in which a control and an experimental condition will be analyzed, triplicates of each biological assay should be performed. After experimentation, each condition should be processed for RNA isolation, array hybridization, data acquisition and normalization, as detailed in the following subheadings. In addition, several other steps are also recommended such as independent validation and assessment of predicted functional roles. Examples of the latter are also illustrated below.
2.1 Isolation of RNA for microRNA Microarrays
Purification and preparation of total RNA that includes small RNAs (3 per condition) into a single quadruplicate containing chip per condition. Within the first condition, 9 reads per microRNA were obtained while within the second approach 4 reads per microRNA were generated. In both cases, appropriate statistical analyses can be performed (>3 reads per condition). Since distinct biological questions were asked on each experiment it is difficult to assess which of the approach was most adequately developed. However, validation analyses of a representative set of those microRNAs identified by microRNA microarray analyses revealed a rather similar validation rate (>80 %). Thus, it is rather likely that running one single microarray per condition leads to similarly robust differentially expressed microRNA identification. We have used two distinct microRNA microarray platforms to dissect differentially expressed microRNAs in distinct biological contexts, mainly cardiac and skeletal muscle development. On the one hand we used N-code miRVana arrays (Life Technology) and on the other hand Agilent arrays. Generating microRNA analyses with mirVana arrays required that each biological sample is hybridized to a single array and all arrays (6 in the case of a 2 conditions, 3 replica analyses) are run in parallel. On the other hand, Agilent arrays contained 8 arrays within a single glass and therefore hybridization, probe clearance and signal scanning were always run simultaneously. In our experience, both platforms were successfully used and
Diego Franco et al.
identification of differentially expressed microRNAs was achieved in both cases. However, it is important to highlight that miRVana arrays displayed larger data variations as compared to Aligent arrays, yet miRVana arrays were more versatile for experimentation design. Our data are in line within a recent report by Callari et al. (16) since in several cases, discordant results are obtained within similar physiopathological conditions, suggesting that a large variability on data acquisition and analyses. These authors compared four distinct microarray platforms (Agilent, Exiqon, Illumina, Miltenyi) within the same biological context, colon cancer tissues. They found a poor overlap among differentially expressed genes. Interestingly, those differentially expressed microRNA with high concordant correlation among distinct platform where equally validated by qRT-PCR. Thus, these data suggest that independent of the selected platform, validation by qRT-PCR is compulsory. In any case, after choosing the appropriate platform and experimental design, arrays hybridization, clearance and input screening was externalized to a national microarray analyses plat˜ a, Madrid) and more recently to a commercial form (Genoma Espan SME (Bioarrays, SL, Alicante). Hybridization conditions, clearance and input screening were performed by specialized staff and raw data were obtained. To our point of view this is rather convenient for small to medium academic institutions in which acquiring and using microarray analyses platform is rather unaffordable.
4
microRNA Data Analysis, Normalization and Representation In microRNA profiling experiments, using microarray technology, an adequate analysis has to be achieved in order to avoid incorrect conclusions. As in mRNA gene expression microarray procedures, the data analysis pipeline usually includes preprocessing, normalization, parametric or nonparametric statistical analysis to detect those microRNAs differentially expressed in our experimental model, multivariate data exploration and gene enrichment functional analyses. We also follow this well established pipeline in our microRNA profiling studies in cardiogenesis, which were carried out using free software as described below. In general, microRNA microarray preprocessing and statistical analyses are performed calling Bioconductor functions (bioconductor.org/) in R software (r-project.org). 1. Once the raw data are obtained (after background correction, which depends on platform), we usually impute those densitometry values
Analysis of microRNA Microarrays in Cardiogenesis Diego Franco, Fernando Bonet, Francisco Hernandez-Torres, Estefania Lozano-Velasco, Francisco J. Esteban, and Amelia E. Aranega Abstract microRNAs are a subclass of noncoding RNAs which have been demonstrated to play pivotal roles in multiple cellular mechanisms. microRNAs are small RNA molecules of 22–24 nt in length capable of modulating protein translation and/or RNA stability by base-priming with complementary sequences of the mRNAs, normally at the 30 untranslated region. To date, over 2,000 microRNAs have been already identified in humans, and orthologous microRNAs have been also identified in distinct animals and plants ranging a wide vast of species. High-throughput analyses by microarrays have become a gold standard to analyze the changes on microRNA expression in normal and pathological cellular or tissue conditions. In this chapter, we provide insights into the usage of this uprising technology in the context of cardiac development and disease. Keywords: microRNA, Microarrays, Cardiac development, Meta-analyses
1
Background Cardiac development is a complex process in which multiple cell types are involved (1–3). From the early stages of cardiogenic specification soon after gastrulation, the heart progressively acquires a tubular shape formed by an inner endothelial lining and an outer myocardium layer (4). The heart is the first organ to display left-right asymmetry (5, 6) and more importantly, it is the first organ to be functional during organogenesis (2). Soon after the heart start to pump, atrial and ventricular chambers are progressively configured and valve primordia are formed in the intertwining areas, i.e., outflow tract, atrioventricular canal, and inflow tract (2). Subsequently, the heart is separated and each of these structures becomes divided into left and right parts (7, 8). Over the last decades, we have gained crucial insights into the molecular mechanisms that govern cardiac morphogenesis. Given the complexity of cardiac morphogenesis, multiple pathways and transcriptional factors are involved during cardiogenesis at different stages, as recently reviewed (9, 10). In the last years, a novel layer of complexity is emerging in the cardiovascular
Diego Franco et al.
development field, namely the post-transcriptional regulatory networks driven by noncoding RNAs. Among noncoding RNAs, microRNAs have emerged as a pivotal mechanism is cardiovascular development, since targeted deletion of the microRNAs processing ribonuclease Dicer resulted in severe cardiovascular defects. Moreover, targeted deletion of single microRNAs, such as miR-1 and miR-126, respectively, also resulted in severe cardiovascular development impairment (11, 12). In view of the important contribution of distinct microRNAs in cardiovascular field, great efforts have been devoted to map the microRNA microarray fingerprints of distinct normal and abnormal cardiovascular contexts (see for a recent review (13–15)). In this chapter, we provide insights into the usage of microRNA microarray analyses in the field of cardiovascular development, as well as insights as how to proceed further beyond the classical microarray approach, such as microarray meta-analyses.
2
Materials and Methods Initial steps to ensure appropriate microRNA microarray analyses start already on the experimental design. Considering a simple experiment in which a control and an experimental condition will be analyzed, triplicates of each biological assay should be performed. After experimentation, each condition should be processed for RNA isolation, array hybridization, data acquisition and normalization, as detailed in the following subheadings. In addition, several other steps are also recommended such as independent validation and assessment of predicted functional roles. Examples of the latter are also illustrated below.
2.1 Isolation of RNA for microRNA Microarrays
Purification and preparation of total RNA that includes small RNAs (3 per condition) into a single quadruplicate containing chip per condition. Within the first condition, 9 reads per microRNA were obtained while within the second approach 4 reads per microRNA were generated. In both cases, appropriate statistical analyses can be performed (>3 reads per condition). Since distinct biological questions were asked on each experiment it is difficult to assess which of the approach was most adequately developed. However, validation analyses of a representative set of those microRNAs identified by microRNA microarray analyses revealed a rather similar validation rate (>80 %). Thus, it is rather likely that running one single microarray per condition leads to similarly robust differentially expressed microRNA identification. We have used two distinct microRNA microarray platforms to dissect differentially expressed microRNAs in distinct biological contexts, mainly cardiac and skeletal muscle development. On the one hand we used N-code miRVana arrays (Life Technology) and on the other hand Agilent arrays. Generating microRNA analyses with mirVana arrays required that each biological sample is hybridized to a single array and all arrays (6 in the case of a 2 conditions, 3 replica analyses) are run in parallel. On the other hand, Agilent arrays contained 8 arrays within a single glass and therefore hybridization, probe clearance and signal scanning were always run simultaneously. In our experience, both platforms were successfully used and
Diego Franco et al.
identification of differentially expressed microRNAs was achieved in both cases. However, it is important to highlight that miRVana arrays displayed larger data variations as compared to Aligent arrays, yet miRVana arrays were more versatile for experimentation design. Our data are in line within a recent report by Callari et al. (16) since in several cases, discordant results are obtained within similar physiopathological conditions, suggesting that a large variability on data acquisition and analyses. These authors compared four distinct microarray platforms (Agilent, Exiqon, Illumina, Miltenyi) within the same biological context, colon cancer tissues. They found a poor overlap among differentially expressed genes. Interestingly, those differentially expressed microRNA with high concordant correlation among distinct platform where equally validated by qRT-PCR. Thus, these data suggest that independent of the selected platform, validation by qRT-PCR is compulsory. In any case, after choosing the appropriate platform and experimental design, arrays hybridization, clearance and input screening was externalized to a national microarray analyses plat˜ a, Madrid) and more recently to a commercial form (Genoma Espan SME (Bioarrays, SL, Alicante). Hybridization conditions, clearance and input screening were performed by specialized staff and raw data were obtained. To our point of view this is rather convenient for small to medium academic institutions in which acquiring and using microarray analyses platform is rather unaffordable.
4
microRNA Data Analysis, Normalization and Representation In microRNA profiling experiments, using microarray technology, an adequate analysis has to be achieved in order to avoid incorrect conclusions. As in mRNA gene expression microarray procedures, the data analysis pipeline usually includes preprocessing, normalization, parametric or nonparametric statistical analysis to detect those microRNAs differentially expressed in our experimental model, multivariate data exploration and gene enrichment functional analyses. We also follow this well established pipeline in our microRNA profiling studies in cardiogenesis, which were carried out using free software as described below. In general, microRNA microarray preprocessing and statistical analyses are performed calling Bioconductor functions (bioconductor.org/) in R software (r-project.org). 1. Once the raw data are obtained (after background correction, which depends on platform), we usually impute those densitometry values
