Detection of Human micro RNAs across mi RNA

Detection of Human micro. RNAs across mi. RNA Array and Next Generation DNA Sequencing Platforms The 2008 -09 Joint MARG/DSRG Research Group Project Herbert Auer 1, Christina Harrington 2, Susan Hester 3, Wei Wang 4, Steve Potter 5, Don Baldwin 6, Jay Tiesman 7, Nadereh Jafari 8, Nancy Denslow 9, Peter A. Schweitzer 4, Lalit Ponnala 4, Ken Dewar 10, Jan Kieleczawa 11, Robert Steen 12, Michael Zianni 13, Doug Bintzler 14, Anoja Perera 15, Sushmita Singh 16, and Michelle M Detwiler 17 1 Institue for Research in Biomedicine, Barcelona, Spain; 2 Oregon Health & Science University, Portland, OR; 3 Environmental Protection Agency, Research Triangle Park, NC; 4 Cornell University, Ithaca, NY; 5 Cincinnati Children’s Hospital, Cincinnati, Ohio; 6 University of Pennsylvania, Philadelphia, Penn; 7 Procter & Gamble, Cincinnati, OH; 8 Nortwestern University, Chicago, Il; 9 University of Florida, Gainesville, FL; 10 Mc. Gill University, Montreal, Quebec; 11 Wyeth Research, Cambridge, MA; 12 Harvard Medical School, Boston, MA; 13 Ohio State University, Columbus, Ohio; 14 DNA Analysis, Inc. , Cincinnati, OH; 15 Stowers Institute for Medical Research, Kansas City, MO; 16 University of Minnesota, St. Paul, MN; 17 Roswell Park Cancer Institute, Buffalo, NY micro. RNA (mi. RNAs) are non-coding RNA molecules between 19 and 30 nucleotides in length that are believed to regulate approximately 30 per cent of all human genes. Recent developments in microarray technologies and the introduction of high throughput DNA sequencing (HT DNA Seq) now make it possible to use these advanced platforms for mi. RNA-expression profiling. To determine the effectiveness of each of these platforms in measuring mi. RNA expression and to compare the accuracy of the microarray data. HT DNA Seq profiles and quantitative RT-PCR results, the Microarray Research Group (MARG) and the DNA Sequencing Group (DSRG) developed a joint research project. The goal of the MARG component of the research project was to evaluate mi. RNA platforms for their ability to detect mi. RNAs from complex total RNA samples. In addition to 3 DNA microarray platforms, two PCR-based platforms were included in the study for performance comparison. Each of the 5 mi. RNA platforms was tested with total RNA to evaluate the ability of each platform to detect and measure individual mi. RNAs in a complex biological sample. Aliquots of RNAs from two human tissue were analyzed at separate test sites for each of the following mi. RNA platforms: Agilent mi. RNA microarray, Illumina mi. RNA expression panel, Exiqon mi. RCURY LNA arrays, Applied Biosystems Taq. Man mi. RNA assay, and Affymetrix Gene. Chips. The second component of the study was performed by the DSRG. mi. RNA expression profiles of the same two RNAs were determined using the “Next Generation” HT DNA sequencer, Illumina Genome Analyzer (Solexa). The mi. RNA profiles produced by the HT DNA sequencer were compared to mi. RNA array results and to RT-PCR data. The results of these comparisons are presented. EXPERIMENTAL DESIGN Multiple microarray platforms for measuring micro. RNA (mi. RNA) patterns in total RNA samples were compared with each other and with Taqman Low Density Arrays (TLDA) and Solexa DNA sequencing. All assays were performed on commercial total RNA from a single source using manufacturer recommended protocols. 3 replicate assays were performed for each sample on all platforms except for DNA sequencing. Data was collected centrally and analyzed. Platform Array Triplicates have been performed for each of the platforms. From the average of replicates, the number of mi. RNAs were counted for which each of the platforms measured differential expression > 2 -fold between the two samples (left columns). Subsequently, the number and fraction of differentially expressed mi. RNAs that also showed statistical significance was determined by a t-test (p<0. 05) (right columns). The more technical variability within a platform, the higher the percentage of mi. RNAs with > 2 -fold differential expression but without statistical significance (triangles). Correlation across platforms The Pearson correlation coefficient was calculated for the average measurement of differential expression of each platform relative to the other ones. Table 2 shows these correlations. Agilent 3 mi. RNA_v 1 Gene. Chip Agilent 1 250 ng 3 Human mi. RNA microarray Exiqon Applied (ABI) 2 500 ng 3 Taq. Man Human mi. RNA A & B Array v 2. 0 Exiqon 3 800 ng 3+3* mi. RCURY LNA™ mi. RNA Array Illumina Table 1 200 ng 3 Human mi. RNA Expression Profiling v 2 Panel Illumina 1 5000 ng none Genome Analyzer II ABI TLDA arrays: Total RNA was reverse-transcribed with mi. RNA-specific primers, followed by real-time PCR with Taq. Man probes. Each sample was assayed with an A & B card= total 667 mi. RNA measured (each in duplicate). The Taq. Man Human Micro. RNA A Array v 2. 0 contains 384 Taq. Man Micro. RNA Assays enabling quantitation of 377 human micro. RNAs and the Taq. Man Human Micro. RNA B Array v 2. 0 contains 384 Taq. Man Micro. RNA Assays enabling quantitation of 290 human micro. RNAs. Three Taq. Man Micro. RNA Assay endogenous controls are included with each card to aid in data normalization and one Taq. Man Micro. RNA Assay not related to human is included as a negative control. The assays target only mature micro. RNAs. Illumina (Solexa) sequencing: Small RNAs were purified from total RNA using size selection on an acrylamide gel. After ligation of an adapter to the 5’ end, ligated products were size selected again on an acrylamide gel. Subsequently, the same process was repeated ligating an adapter to the 3’ end. After c. DNA synthesis and 15 cycles of amplification, amplified c. DNA was gel-purified again to generate the small RNA libraries. Clonal amplification was performed using one lane per sample in the Cluster station. 36 cycles of sequencing were performed on the Genome Analyzer II. Illumina arrays: Total RNA was polyadenylated using Poly-A- polymerase. Biotinylated oligo-d. T primer with a Universal PCR sequence at its 5’ end was used to make labeled c. DNA. The biotinylated c. DNA was hybridized to micro. RNA-specific oligos and the mixture was bound to streptavidin-conjugated paramagnetic particles to select the c. DNA/oligo complexes. After the oligo annealing, miss-hybridized and non-hybridized oligos were washed away. DNA polymerase was used to extend the specific primer. The extended products were eluted and after PCR amplification, the labeled, single-stranded products were hybridization to the beadchip overnight at 45 C. Fluorescence intensity was measured by the Bead. Array Reader. The human Micro. RNA expression profiling V 2 panel on the 12 sample Beadchip was used for this project. The human v 2 mi. RNA panel contains 1, 146 assays, for detecting more than 97% of the mi. RNAs described in the mi. RBase database. Applied Exiqon Illumina Affymetrix 1 0. 64 0. 81 0. 64 0. 69 1 0. 66 0. 74 0. 76 1 0. 65 0. 72 1 0. 74 Affymetrix 1 Reproducibility across platforms mi. RNAs were counted where 2, 3, 4 or all 5 platforms provided for brain vs. liver > 2 -fold differential expression with statistical significance (p<0. 05). The better the cross-platform comparability of a certain array, the more of its measurements should be confirmed by other types of arrays. Table 3 shows counts of all comparisons. Agilent arrays: Total RNA was labeled with Agilent mi. RNA Complete Labeling and Hyb Kit in which cyanine 3 -cytidine bisphosphate (p. Cp) reagent selectively labels and hybridizes mature mi. RNAs. Samples were hybridized with the Agilent Human mi. RNA Microarray 8 x 15 k version 2 arrays which represents 723 human and 76 human viral mi. RNAs. The arrays were washed and scanned on Agilent array scanner to generate Gene. View files for further analysis. Exiqon arrays: The Exiqon system is a standard dual-channel, glass slide array scanned on an Axon Gene. Pix scanner – for this study we used three replicates of each RNA type with dye swaps. The. gpr files contain the raw and background corrected signals, and the various columns are described in the experiment documents. We used loess normalization, flag signals that are less than 2 SD above background, and average Hy 3 -Hy 5 channels for each sample before running SAM FDR test for significant differences between conditions. -10. 0% Agilent Applied Table 2 * dye-swaps Affymetrix arrays: 1. 5 ug of total RNA was labeledusing the 3 DNA Array Detection Flash Tag RNA Labeling Kit(Genishere), according to manufacturer recommendations. First, poly(A) tailing is carried out at 37 C for 15 min in a volume of 15 ul reaction mix, which contains 1 X Reaction Buffer, 1. 5 ul 25 m. M Mn. Cl 2, 1 ul 1: 500 diluted ATP Mix and 1 ul PAP enzyme. Second, Flash. Tag Ligation is performed at room temperature for 30 min by adding 4 ul of 5 X Flash. Tag Ligation Mix Biotin and 2 ul T 4 DNALigase into the 15 ul of reaction mix. To stop the reaction 2. 5 ul of Stop Solution is added. Samples were hybridized, washed and scanned with an Affymetrix Scanner. -6. 8% To learn about comparability of results across platforms, a list of mi. RNAs measured commonly across all five platforms was extracted. For 646 mi. RNAs, all four platforms provided probes and therefore measurements. For these commonly measured mi. RNAs, correlation across platforms was determined, as well as reproducibility across platforms and sensitivity of measurements. 1500 ng 4 -33. 0% Comparability of measurements across platforms 5 Illumina -25. 9% -8. 3% Fig. 3 Affymetrix METHODS Aliquots of two commercial total RNAs (Ambion First. Choice® Human Total RNA from liver and brain) were distributed to each testing site from a central core laboratory. Each site processed the RNA samples according to manufacturer recommendations for the selected platform. Site Input RNA Replicates Reproducibility within platforms Table 3 Sensitivity of platforms for shared measurements For the 646 mi. RNAs commonly measured on all of the platforms, the number of significant measurements of over 2 -fold differential expression was determined for each of the platforms (Fig. 4). Fig. 4 Results from Sequencing mi. RNA on the Illumina Genome Analyzer II (Brain mi. RNA sample) RESULTS Coverage of platforms The total number mi. RNAs in the human transcriptome is not determined yet and each of the platforms provides a different degree of coverage. Fig. 1 shows the number of human mi. RNAs measured on each of the platforms. Fig. 5 Fig. 1 Summary of results Sensitivity of platforms As a parameter of sensitivity of platforms, the number of mi. RNAs measured as over 2 -fold differentially expressed between brain and liver was evaluated and is also shown as percentage of all measured mi. RNAs (Fig. 2 left and right respectively). Fig. 2 The interrogated platforms for measurements of mi. RNA expression provide different advantages and disadvantages. This study highlights many of these properties and should help to decide which platform suits best for a certain application: èIllumina arrays provided measurements for the highest number of human mi. RNAs and shows highest sensitivity for differential expression èAffymetrix arrays showed highest technical reproducibility èExiqon and Agilent arrays showed best correlation of results èAgilent arrays showed highest sensitivity for mi. RNAs measured on all platforms and highest reproducibility of results by the other platforms èNext Generation Sequencing of the identical RNAs is an ongoing project and additional results will be published