AML engraftment in the NOD/SCID assay reflects the outcome of AML: implications for our understanding of the heterogeneity of AML
Blood Pearce et al. 107: 1166
RNA extraction, small sample cRNA target preparation, and hybridization
Total RNA was extracted from a total of 50 × 103 to 200 × 103 thawed cells using the Trizol purification method (Invitrogen, Paisley, United Kingdom). Before further using the RNA, quality and integrity of RNA was checked on an Agilent Bioanalyzer (Agilent Technologies, Stockport, Cheshire, United Kingdom). Measurement of the quantity of RNA was done using a NanoDrop Spectrophotometer (NanoDrop, Wilmington, DE). Two rounds of amplification were carried out in order to produce first-strand and second-strand cDNA synthesis. Amplification was performed on 100 ng starting material of total RNA, with the Superscript ds-cDNA Synthesis Kit (Invitrogen). First strand synthesis was done by using 5 µM T7-(T)24 primer in the first round (HPLC purified), followed by the use of 0.2 µM random primers in the second round of amplification. Amplified biotinylated complementary RNA (cRNA) was produced with an in vitro transcription labeling reaction, performed according to the manufacturer’s recommendation (Enzo Diagnostics, New York City, NY). Samples with a yield greater than 40 g cRNA were subsequently fragmented and hybridized to Affymetrix U133A oligonucleotide arrays (Affymetrix, Santa Clara, CA) at 45°C for 16 hours. Arrays were washed and stained with streptavidin-phycoerythrin (SAPE; Molecular Probes, Eugene OR). Signal amplification was performed by using a biotinylated antistreptavidin antibody (Vector Laboratories, Peterborough, United Kingdom) following the Affymetrix protocol for high-density chips. Scans were carried out on a GeneArray scanner (Agilent Technologies), and the fluorescence intensities of scanned arrays were analyzed with Affymetrix MicroArray Suite 5.0 (MAS 5.0) software.
Microarray statistical analysis
Affymetrix MAS 5.0 was used for the quantification of gene expression levels. Global scaling was applied to the data to adjust the average recorded to a target intensity of 100. Quantification data were then exported from MAS 5.0 into the R software statistical package (available free of charge at http://www.r-project.org) for further analysis.1 From the 22 283 probes on the Affymetrix U133A chip, we identified 8349 that were present in all the samples investigated. We normalized our data via quantile normalization. This method assumes that the distribution of genes is nearly the same in all chip samples. To normalize each chip, we then computed, for each value, the quantile of that value in the distribution of probe intensities, then transformed the original value to that quantile’s value on the pooled distribution of probes on all chips.
Genes with minimal variation across the samples being analyzed were excluded by subjecting gene expression values to a variational filter that allowed the exclusion of genes when maximum/minimum ratio of normalized expression values was less than 5. As a result, 2476 probes and 2260 probes remained for quantile normalization 2260 for median normalization, respectively. This allowed us to exclude noninformative and nonchanging genes from the 8349 probes across all 8 chips.
Unsupervised hierarchical clustering was performed on the samples, with Euclidean distance as a metric and complete linkage for the clustering algorithm. This analysis was applied to the normalized data with or without the variational filter. The cluster dendrogram gave similar results with both lists of genes in either situation.
A statistical group comparison approach (t test) was used to identify genes with statistically significant differences in expression levels between groups of samples.
1. Hornik K. The R Faq. http://www.r-project.org. Accessed on November 16, 2005.