To the editor:

Mahadevappa and Warrington1 developed a protocol that significantly improves sample preparation for gene expression using high-density DNA arrays. The method for expression analysis with high-density arrays includes RNA extraction, double-stranded cDNA synthesis, and production of cRNA by in vitro transcription before the sample is ready for hybridization. Until now, the method has required isolation of mRNA as template for the cDNA synthesis. Because of sample loss during the poly(A) RNA extraction, it has been necessary to use large amounts of cells or tissue as starting material. A major advantage of the new protocol is the possibility of using total RNA, instead of mRNA, as the template for cDNA synthesis1. By eliminating the poly(A) purification step, sample loss is markedly reduced, making the technique useful for analysis of tissues that are impossible to obtain in large quantities.

For preparation of total RNA, Mahadevappa and Warrington used affinity resin extraction (RNeasy; Qiagen, Chatsworth, CA). In our hands, this method consistently results in lower yield compared with other methods for preparation of total RNA. Furthermore, when total RNA, extracted according to Chomczynski and Sacchi2, was further purified using RNeasy, we routinely noted a loss of approximately 30% of the RNA. Gel electrophoresis indicated that the loss of RNA mainly was due to loss of small RNAs. We tested the capability of the RNeasy system to process tRNA, which accounts for a large proportion of small RNAs in the cell, and found that more than 80% of the tRNA was lost. This suggests that the RNeasy system relatively selectively extracts mRNA and ribosomal RNA, while most of the small RNAs, including tRNA, which accounts for approximately 20% of all RNA, are lost.

It is well known that the relative amount of mRNA varies greatly depending on the method used for poly(A) extraction. However, previous methods used for extraction of total RNA have had little influence on the relative abundance of specific RNA molecules and thus allowed direct comparison between methods. The work by Mahadevappa and Warrington clearly facilitates gene expression monitoring from a small number of cells. However, it should be clarified that the total RNA used was not comparable to total RNA extracted by other methods.

The altered proportion of different RNAs could be important for several reasons. The specific loss of small RNAs when using the RNeasy method implies that the relative amount of mRNA in the total RNA preparation is increased. Therefore, the amount of total RNA needed for expression analysis may have to be increased if methods other than RNeasy are used. In addition, other parts of the protocol that were optimized by Mahadevappa and Warrington—for example, the ratio of reverse transcriptase to total RNA or the temperature for the first-strand cDNA synthesis—may need to be adjusted depending on the method used for total RNA preparation. It is also possible that the higher amount of tRNA in total RNA prepared by methods other than RNeasy could affect the efficiency of the cDNA synthesis.

In summary, the protocol recently developed by Mahadevappa and Warrington is a significant improvement, and it will greatly facilitate the use of DNA-arrays for expression analysis in small samples. However, before the protocol can be used for total RNA extracted by conventional methods, it may need further validation.