CASE STUDY: Network Analysis of micro RNA using MetaCore Background MicroRNAs (mirnas) are approximately 22-nucleotide long non-coding RNAs that regulate (by silencing) gene expression through base pairing with target messenger RNAs (mrna). MiRNAs begin as transcribed long RNA precursors and are processed in the nucleus by the RNase III enzyme complex that includes Drosha-Pasha/DGCR8 to generate approximately 70-base pre-mirnas. Pre-miRNAs are processed further as they are incorporated into an Argonaute- containing RNA-induced silencing complex (RISC), where mirnas pair to their targets through base-pairing with the 3 -untranslated region. As a result translation of the mrna target is repressed primarily through cleavage mechanisms (1,2). Although mirna were first described in Caenorhabditis elegans (3), thousands of mirnas have been identified in several organisms. Intriguingly, the biological functions of most mammalian mirnas are still poorly understood but have been found to play a role in embryogenesis and stem cell maintenance, hematopoietic cell differentiation (1,4), and brain development (5). Notably, most of these processes, when mis-regulated, lead to cancer (1, 6, 7). It is not a surprise therefore that mirnas are now emerging as cancer targets. Notably, human mirnas in particular, are often found in genome regions that are commonly amplified or deleted in human cancer (8). Also, when compared to normal tissues, malignant tumors and tumor cell lines have widespread deregulated mirna expression (9, 1). Consequently, identifying patterns of mirna expression in disease states and elucidating the processes dictated by their targets will be key in defining mechanisms of disease progression or classification. Furthermore, the push to use mirna in cancer diagnostics is supported by the fact that mirna is more stable than mrna and can withstand rigorous analysis protocols. Here we use a MetaCore report on the analysis of a mirna expression profile from primary human breast tumors to identify trends in mirna expression in cancer subtypes and further elucidate functional processes associated with each. Figure 1(A). Auto-expand network of mirnas that increase in (mirna 93, 19a, 210, 106b, 25). Predicted targets for each are highlighted in pink, while known are in blue. Notice all downstream interactions from the mirnas are inhibitory (red lines), indicative of mirna function. Materials and Methods The publically available data used in this case study was obtained from the Gene Expression Omnibus (GEO). Specifically, mirna expression data (GSE7842) was submitted by Blenkiron et al. (10) and the subtype gene expression data (GSE4382) by Sorlie et al (11). Predicted targets of the selected mirnas were obtained from http://microrna.sanger.ac.uk.
Blenkiron et al. used bead-based flow cytometry coupled hybridization in combination with Luminax array technology to obtain expression profiles of 309 mirnas from 93 human breast tumors. Their objective was to use mirna expression for classification of breast tumor subtypes (Luminal A, Luminal B, Basal-like, HER2+ and Normal-like). They determined that the differences in mirna expression observed are mostly likely due to the regulation of mirna expression at the transcriptional and post-transcriptional level and not the loss of gene expression. We calculated the average expression of each mirna from each subtype for analysis in MetaCore. Sorlie et al. reported characteristic patterns of gene expression classifiers of breast tumors into clinically relevant subgroups. For this case study, we selected three samples from the Erbb2 group and three from the Basal subtype to overlay onto the mirna networks generated with the Blenkiron et al. profiles and predicted targets. For a more detailed description of MetaCore analysis of the full Sorlie data set contact training@thomsonreuters.com. We used MetaCore to build networks of the mirnas and identified trends of mirna expression within each subtype or tumor grade status (Table 1). We then built expanded (auto expand, 200 nodes) networks for each trend with known and predicted targets for each mirna involved, using interaction target files (refer to MetaCore help section for details on how to create and upload an interaction file). Enrichment analysis was performed using exported lists from each trend network to compare/determine top functional processes in identified trend. Lastly, we overlaid expression data from Basal and Her2+ breast cancer subtypes to demonstrate a causative relationship with changes mirna expression and target expression from tumor samples. Results By building networks of calculated average mirna expression, according to subtype or grade, we determined subsets of mirnas carried similar patterns of expression for each designated trend (Table 1). For instance, mirnas that increase with grade progression include: 106b, 25, 210, 19a, 93, and mirnas that decrease with grade progression include 130a, let-7e. MiRNAs highly increased throughout all subtypes include 148b, 34a and decrease throughout include 23a, 3. Interestingly trends of mirna that increase in both Basal and Her2+ (mirna 106b, 25) decrease in both Luminal A and B and those that decrease in Basal and Her2+ (mirna 130a, let 7e, 497) increase in Luminal A and B. Trends specific to each subtype can be identified as well ( see Table 1). Notable, the trends associated with mirna 106b, 25, 210, 19a, 93, and 130a were also identified in the original study. mirna Subtype Trends Grade Trends mirna Basal Her2+ Luminal A Luminal B 497 497 let7e let7e 130a 130a 106b 106b 25 25 19a 19a 93 93 210 210 141 141 375 375 149 149 345 345 34a 34a 148b 148b 23a 23a 31 31 Table 1: mirna expression trends identified using MetaCore within each breast cancer subtype or grade progression. Using the identified trends, we chose to build networks (auto-expand, 200 nodes) for each trend with the representative mirnas and activated experiment and interaction files of the predicted targets for each. For instance, for the increase in grade progression, the network was built with mirnas106b, 25, 210, 19a, 93. Samples of these networks are shown in Figures 1 and 2, and highlight known identified targets (from the MetaCore data base) in blue lines and additional predicted targets (from interaction file) in pink lines.
Figure 1 (B). Auto-expand network of mirna that decrease in (mirna 130a and Let-7e). Predicted targets for each are highlighted in pink, while known are in blue. Notice all downstream interactions from the mirnas are inhibitory (red lines), indicative of mirna function. Figure 2(A). Auto-expand network of mirna that increase in expression (mirna 106b, 25) in aggressive subtypes of breast cancer (Basal and Her2+). Predicted targets for each are highlighted in pink, while known are in blue. Notice all downstream interactions from the mirnas are inhibitory (red lines), indicative of mirna function. Figure 2(B). Auto-expand network of mirna that increase in expression (mirna 497, Let-7e) in Luminal A and B subtypes of breast cancer. Predicted targets for each are highlighted in pink, while known are in blue. Notice all downstream interactions from the mirnas are inhibitory (red lines), indicative of mirna function.
To determine biological functions represented by each trend, we exported the gene lists associated with each network shown above, and used MetaCore to create experimental or GX files that can be used for function enrichment analysis. Figure 3 represents the top processes according to GeneGo Map Folder categories for each trend. GeneGo Map Folders are organized to collect all GeneGo Maps that describe different canonical pathways involved in the same process. For instance, G1 and G2 transition may be separate maps but are both part of cell cycle regulation and would be in the cell cycle GeneGo Map Folder. Interestingly, the most significant GeneGo Map Folder (Cell cycle and its regulation) for mirna processes that increase with grade progression is shared by mirna processes elevated in Luminal A and B. Likewise, the most significant GeneGo Map Folder (Vascular Development-Angiogenesis) for mirna processes that decrease with grade progression is shared by mirna processes elevated in Basal and Her2+. Consequently using the predicted targets incorporation allowed further expansion (included more targets and a larger comprehensive list to analyze) of each network that ultimately resulted in the ability to identify processes unique to each trend. For instance, in Figure 3A, inflammation as well as tissue remodeling is only represented by the network associated with mirnas that increase in grade progression. Diuresis is exclusive to the decrease with grade progression trend, while vasoconstriction, neurotransmission, and calcium signaling is only observed in mirna networks associated with Basal and Her2+ subtypes. Lastly, mirna trends representing Luminal A and B trends describe androgen signaling. A: Increase with Grade Progression B: Decrease with Grade Progression C: Increase with Basal/ Her2+ D: Increase with Luminal A and B Figure 3: Histograms representing top GeneGo Map Folders from the enrichment analysis of the gene lists associated with the trends identified in Table 1. Key differentially representative processes are represented in rectangles. In each histogram, the longer the length of each bar the more significant (-log pvalue) the folder. The results are organized according to descending significance. In attempt to connect the trends identified above with known expression data, we overlaid gene expression from Basal and Her2+ breast cancer samples (Sorlie et al.) on our mirna target networks. Note that objects that are expressed and decreased in expression are marked with blue circles while those that are increased are marked with red circles in the upper right hand corner of each. We hypothesized that an increase in mirna expression should lead to a decrease in target expression. We can then use the targets that have a decrease in expression as markers for that subtype, in association with the mirna. As shown in Figure 4, p21 and MTF are targets of mirna 106b that decrease in expression in the Basal subtype (Sorlie et. al.), but PTEN is lost in the Her2+ subtype (Sorlie et al). The loss of target expression from each Sorlie data set is summarized in Table 2. Furthermore, using the Sorlie et al. data we determined that most Her2+ targets were associated with Let-7e while targets expressed in the Basal subtypes were associated with a variety of mirnas (487, 21, 130a).
Figure 4 (A): Auto-expand network of mirna that increase in with predicted and known targets (similar to Figure 1A). Expression data from tissue samples of Basal subtype (taken from Sorlie et al.) is overlaid (blue or red filled circlestop right of selected objects). Targets that decrease in expression (targets with blue circles), in accordance with increase mirna expression are highlighted (open red circles). Figure 4 (B): Auto-expand network of mirna that increase in with predicted and known targets (similar to Figure 1A). Expression data from tissue samples of Her2+ subtype (taken from Sorlie et. al.) is overlaid (blue or red filled circlestop right of selected objects). Targets that decrease in expression (targets with blue circles), in accordance with increase mirna expression are highlighted (open red circles). 93 210 25 19a 106b Loss of Expression of Identified Targets mirna with grade Progession * mirna In Basal/ Her2+ mirna with grade Progession ** mirna In Luminal A/B Solie Basal Sorlie Her2+ Solie Basal Sorlie Her2+ Solie Basal Sorlie Her2+ Solie Basal Sorlie Her2+ p21, MTF PTEN p21 CTGF, SREBP1, THSD1 CTGF, SREBP1, AP1 BTG2, SREBP1 BTG2, MDM2 BTG2 BTG2 SREBP1, SREBP1, TCF4, InsP6Kinase, GluR2 GluR2 PTEN, p14arf let 7e 30a 13 497 SelenBP, eif3s1, MHC Class I CSF1R, Calnexin, MafB DNA Polymerase, p21, eif3s1, MHC, ACDS, MDM2, IGF1R, CEP110, SelenBP * Similar trends identified in Luminal A/B **Similar trends identified in Basal/ Her2+ CSF1R, Calnexin, MafB SelenBP, SIAH, PACT DNA Polymerase, ACDS, MDM2, IGF1R, p14arf Table 2: Summary of observations obtained by overlaying gene expression data from cancer subtypes on mirna target networks (identified above). Networks for each mirna trend (Figure 1) were overlaid with expression data from Sorlie et al. (basal and Her2+). Targets that were found to be lost in expression were noted for each mirna in each network and tabulated below. Conclusions: The Blenkiron et al. study provided a basis for mirna expression analysis as a tool in breast cancer classification. Here we expanded on their conclusions and identify groups of mirnas based on their expression profiles, network and pathway analysis to facilitate the identification of the grade, and/or subtype of breast cancer. By expanding the mirna network space with known and predicted mirna targets, we also identified key processes represented by each trend. Furthermore, expanded mirna networks can be used in combination with expression analysis to delineate mirna targets specific for each subtype. MetaCore and the interaction file function analysis tools were key in: 1) analyzing changes in expression (trends) of mirna in each subtype, 2) depicting mirna target networks, and 3) describing functional processes associated with each mirna network (known and predicted).
References 1) Blenkiron C., Miska E.A. (2007) mirnas in cancer: approaches, aetiology, diagnostics and therapy. Human Molecular Genetics 16, R106 R113. 2) Han, J., Lee, Y., Yeom, K.H., Kim, Y.K., Jin, H. and Kim, V.N. (2004) The Drosha DGCR8 complex in primary microrna processing. Genes Dev., 18, 3016 3027. 3) Lau NC, Lim LP,Weinstein EG, Bartel DP (2001). An abundant class of tiny RNAs with probable regulatory roles in Caenorhabditis elegans. Science. 294:858 62. 4) Papagiannakopoulos T, Kosik KS. (2008) MicroRNAs: regulators of oncogenesis and stemness.bmc Med.; 24;6:15. 5) Kuss AW, Chen W. (2008) MicroRNAs in brain function and disease. Curr Neurol Neurosci Rep. May;8(3):190-7. 6) Alvarez-Garcia, I. and Miska, E.A. (2005) MicroRNA functions in animal development and human disease. Development, 132, 4653 4662. 7) Zhang, M. A. Farwell micrornas: a new emerging class of players for disease diagnostics and gene therapy. Cell. Mol. Med. Vol 12, No 1, 2008 pp. 3-21 8) Calin, G.A., Dumitru, C.D., Shimizu, M., Bichi, R., Zupo, S., Noch, E., Aldler, H., Rattan, S., Keating, M., Rai, K. et al. (2002) Frequent deletions and down-regulation of micro-rna genes mir15 and mir16 at 13q14 in chronic lymphocytic leukemia. Proc. Natl Acad. Sci. USA, 99, 15524 15529. 9) Lu J, Getz G, Miska EA, Alvarez-Saavedra E, Lamb J, Peck D, Sweet-Cordero A, Ebert BL, Mak RH, Ferrando AA, Downing JR, Jacks T, Horvitz HR, Golub TR. MicroRNA expression profiles classify human cancers. Nature. 2005; 435: 834 8. 10) Blenkiron C, Goldstein LD, Thorne NP, Spiteri I, Chin SF, Dunning MJ, Barbosa-Morais NL, Teschendorff AE, Green AR, Ellis IO, Tavaré S, Caldas C, Miska EA. MicroRNA expression profiling of human breast cancer identifies new markers of tumor subtype. Genome Biol. 2007;8(10):R214. 11) Sorlie T, Tibshirani R, Parker J, Hastie T, Marron JS, Nobel A, Deng S, Johnsen H, Pesich R, Geisler S, Demeter J, Perou CM, Lønning PE, Brown PO, Børresen-Dale AL, Botstein D. Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A. 2003 Jul 8;100(14):8418-23. THOMSON REUTERS Regional Offices North America Philadelphia +1 800 336 4474 +1 215 386 0100 Latin America +55 11 8370 9845 Europe, Middle East and Africa Barcelona +34 93 459 2220 London +44 20 7433 4000 Contact us to find out more about MetaCore or visit thomsonreuters.com/diseaseinsight Asia Pacific Singapore +65 6775 5088 Tokyo +81 3 5218 6500 For a complete office list visit: ip-science.thomsonreuters.com/ contact LS091149 Copyright 2012 Thomson Reuters