Rna Amplification

RNA amplification is needed to generate the 0.5-10 ^g of RNA needed for a typical microarray analysis, starting from the approximately 10-100-pg total RNA/cell that is recovered from a laser-microdissected sample. This is technically challenging, as attested by the multitude of papers on T7 RNA polymerase-based RNA amplification (Phillips and Eberwine, 1996; Pabon et al., 2001; Zhao et al., 2002; Gomes et al., 2003; Li et al., 2003; Polacek et al., 2003; Spiess et al., 2003; Xiang et al., 2003; Goff et al., 2004; Ji et al., 2004; Kamme et al., 2004; Moll et al., 2004; Rudnicki et al., 2004; Schneider et al., 2004). T7 RNA polymerase-based and PCR-based techniques were developed in 1990 (Brady et al., 1990; Van Gelder et al., 1990) and have evolved into several different versions following attempts to improve amplification yield and throughput. More recently, NuGEN's isothermal amplification system added a new amplification principle. There are several commercial kits of the T7 system available today from a variety of vendors, including Affymetrix, Agilent, Ambion, Molecular Devices, Artus-Biotech, Enzo Life Science, Epicentre Technologies, Genisphere, Roche, Stratagene, and System Biosciences. For laser-microdis-sected samples, a two-round system is normally required.

Considering the sensitivity of RNA amplification to variations in reagent quality, it is probably advisable to at least start off with a commercial kit. Might one avoid RNA amplification issues by simple gross dissection of the tissue and a microarray screen with subsequent confirmation of interesting genes by in situ hybridization? Not really, and the explanation reveals one of the limitations of microarrays. While a typical microarray may have a sensitivity of approximately 1/200,000 to 1/500,000, the complexity of a single cell may be 500,000 individual transcripts or higher. Therefore, if a sample is even moderately complex in terms of cell types, there will certainly be a

FIGURE 8.2 Pairwise scatter plots of seven laser-microdissected samples from rat hippocampus CA1. Approximately 100 cells were captured per sample using PixCell2 (Molecular Devices). Extracted RNA was amplified using TargetAmp (Epicentre Technologies) and hybridized to cDNA arrays as described by Kamme et al. (2003). Plots show log2-transformed, quantile-normalized intensity data. Samples are indicated by Roman numerals I-VII. The overall average Pearson's correlation (R2) was .97.

FIGURE 8.2 Pairwise scatter plots of seven laser-microdissected samples from rat hippocampus CA1. Approximately 100 cells were captured per sample using PixCell2 (Molecular Devices). Extracted RNA was amplified using TargetAmp (Epicentre Technologies) and hybridized to cDNA arrays as described by Kamme et al. (2003). Plots show log2-transformed, quantile-normalized intensity data. Samples are indicated by Roman numerals I-VII. The overall average Pearson's correlation (R2) was .97.

substantial number of expressed genes that fall below the detection limit and become false negatives. In essence, the microarray has a limited bandwidth. To illustrate this point, we used microarrays to analyze the expression of tyrosine hydroxylase in microdissected substantia nigra and in total rat brain RNA. Tyrosine hydroxylase is highly expressed in the substantia nigra and gave a high signal-to-noise ratio in the microdissected sample. However, in the whole-brain sample, it was not detectable (Table 8.1), clearly showing the negative effect of increased sample complexity. This effect of sample complexity on microarray data is frequently overlooked when assessing RNA amplification methods. A popular measure of RNA amplification

TABLE 8.1

Signal-to-Background Ratios for Tyrosine Hydroxylase Expression Using Total Rat Brain RNA or Microdissected Substantia Nigra

Sample Signal/Background

Whole rat brain RNA 1

Microdissected substantia nigra 16

Note: Microdissection was done using PixCell2, and micro-array analysis was performed as described by Kamme et al. (2003).

quality is to plot the correlation between unamplified and amplified RNA as an indication of linearity, or maintained representation, of amplification. However, the unamplified data most likely omit a large amount of false negatives and, as such, are in fact biased against rare genes. If the amplification procedure increases the number of detectable genes, one might argue that this is a more informative representation of the sample, even if it is at the expense of a poorer correlation with an unamplified sample. Indeed, it has been reported that RNA amplification improves the sensitivity of microarray analysis as compared with using total RNA (Feldman et al., 2002), likely by reducing the complexity of the RNA mix, as nonpolyadeny-lated RNAs are not efficiently amplified. An analogy would be proteomic analyses, where albumin and other abundant proteins are regularly removed from samples to permit the detection of less abundant proteins by freeing up "bandwidth" of the assay (Georgiou et al., 2001).

Was this article helpful?

0 0

Post a comment