Evaluation of methods for amplification of picogram amounts of total RNA for whole genome expression profiling

Abstract

Background: For more than a decade, microarrays have been a powerful and widely used tool to explore the transcriptome of biological systems. However, the amount of biological material from cell sorting or laser capture microdissection is much too small to perform microarray studies. To address this issue, RNA amplification methods have been developed to generate sufficient targets from picogram amounts of total RNA to perform microarray hybridisation.

Results: In this study, four commercial protocols for amplification of picograms amounts of input RNA for microarray expression profiling were evaluated and compared. The quantitative and qualitative performances of the methods were assessed. Microarrays were hybridised with the amplified targets and the amplification protocols were compared with respect to the quality of expression profiles, reproducibility within a concentration range of input RNA, and sensitivity. The results demonstrate significant differences between these four methods.

Conclusion: In our hands, the WT-Ovation pico system proposed by Nugen appears to be the most suitable for RNA amplification. This comparative study will be useful to scientists needing to choose an amplification method to carry out microarray experiments involving samples comprising only a few cells and generating picogram amounts of RNA.

Figure 1

Quality of input RNA and targets synthesised from different amounts of input RNA. A, Bioanalyzer electrophoretic profile of the diluted Universal Human Reference RNA used as input for all amplifications. This profile corresponds to a classical and non-degraded human RNA with two fine characteristic peaks corresponding to 18S and 28S RNAs. 8.4 corresponds to the RNA Integrity Number (RIN) and reflects the high quality of this RNA B, electrophoretic profiles of cRNA obtained after one- and two-round Affymetrix amplification using 2 μg, and 100 ng of input RNA respectively or water as negative control. C, D, E, F, electrophoretic profiles of cRNA or cDNA obtained using Ambion (C), Arcturus (D), Epicentre (E) and Nugen (F) amplification systems from 250 pg and 500 pg of input RNA, or water as negative control.

Figure 2

Box plots of signal intensities. A, raw data. B, data after MAS5 normalisation.

Figure 3

Statistical analysis of expression level measurements for assessing reproducibility and comparability of amplification chemistries. A, we calculated the Pearson's correlation coefficients between technical replicates (same chemistry, same amount of input RNA, but different laboratories). The correlation values were then averaged for each chemistry (grey bars). We also calculated correlation coefficients between results obtained from 250 pg and from 500 pg of RNA input (same chemistry, but different amount of RNA input) and averaged them for each chemistry in order to evaluate the robustness across quantitative variability of RNA input (black bars). Most Epicentre amplifications did not yield sufficient aRNA to carry out hybridisations, and Pearson's correlation coefficients could therefore not be calculated (*). B, Graphic representation of Pearson's correlation coefficients calculated for each pairwise comparison of all assays.

Figure 4

Quality of cDNA targets synthesised following Nugen protocol from different amounts of input RNA. Bioanalyzer electrophoretic profiles of cDNA targets obtained after the amplification of 50 pg, 100 pg, 250 pg, 500 pg, and 1 ng of total RNA using Nugen chemistry.

Figure 5

Reproducibility of expression data obtained using Nugen chemistry across amounts of input RNA. Pearson's correlation coefficients were computed and represented graphically for each pairwise comparison of assays. Lower correlations were observed for data obtained from smallest amounts of input RNA (50 pg and 100 pg).

Figure 6

Comparison of differential gene expression data using Affymetrix one-round and Nugen chemistries. Targets were generated from RNAs of two cell lines using Affymetrix 1-round IVT and Nugen chemistries. After hybridisation, differential gene expression results were normalised (MAS5) and the expression ratio calculated for each probeset. Logarithm of ratios to the base 2 (Log2 ratios) were computed for each chemistries and graphed(X axis Nugen and Y axis Affymetrix). Differences in the direction of differential expression were only observed for 4% of the probesets (dots located in the upper-left quadrant and lower-right quadrant). Moreover, almost all of those probesets show a low differential expression (absolute value lower than 1, less than 2-fold over or lower expression). The slope of the regression line is 1.033 (p < 10^-16) and it intercepts the Y axis at -0.02 (p < 10^-12). The Pearson's correlation coefficient has been calculated between the expression ratios distribution (R = 0.321).