Letter

Top Letter References

We would like to thank Drs. Hesse and Court for their reading of and comments regarding our recent publication (Chang et al., 2003a). They suggested that an alternative explanation for our data are that differences in mRNA quality among the human liver samples could have resulted in the interindividual variability in CYP2B6, constitutive androstane receptor (CAR¹), and PXR mRNA levels shown in our study. They also suggested that the variability could be substantially reduced by normalizing the data to the mRNA expression of a control gene (e.g., -actin, glyceraldehyde-3-phosphate dehydrogenase). We agree that the quality of the RNA is always a consideration in the interpretation of gene expression data, especially those collected from clinical specimens. However, we do not believe that differences in the quality among the RNA samples accounted for the interindividual variability in the gene expression reported in our study because the integrity of each isolated RNA sample was assessed by gel electrophoresis and no RNA degradation was observed. Interestingly, Rodriguez-Antona et al. (2001) reported a 582-fold variability for CYP1A2 mRNA levels, whereas we obtained a 500-fold variability in our panel of human liver samples in which CYP1A2 mRNA expression was quantifiable (Chang et al., 2003b). Furthermore, the variability was 278-fold in CYP2B6 mRNA levels and 240-fold in CAR mRNA levels, but it was only 27-fold for PXR mRNA expression in the same panel of human liver samples. If RNA degradation was a major issue, it should have affected the expression of the three genes to a similar extent. Drs. Hesse and Court suggested the use of a negative control to rule out RNA degradation as a possible reason for the significant positive correlations obtained in our study (Chang et al., 2003a). We have measured the mRNA levels of CYP1A1, CYP1A2, and CYP1B1 in the same samples (Chang et al., 2003b). CYP1A2 mRNA levels did not correlate with CYP2B6, CAR, or PXR mRNA levels. This indicates that RNA degradation, if present, could not have been the reason for the correlations between CYP2B6, CAR, and PXR mRNA levels. In general, samples with high CYP2B6, CAR, and PXR mRNA levels had low CYP1A2 mRNA levels and vice versa.

The use of an internal control gene to normalize mRNA expression can correct for RNA degradation in the samples, but it does not necessarily reduce the intersample variability. In a study by Rodriguez-Antona et al. (2001), the investigators employed reverse transcription and real-time PCR to determine the mRNA levels of various cytochromes P450 in a panel of 12 individual human liver samples. In that study, the data were normalized to -actin mRNA expression, and the interindividual differences in CYP2B6 mRNA levels were 158-fold. By comparison, we had a 278-fold variability in our panel of human liver samples (Chang et al., 2003a).

It is a common practice to use the levels of a housekeeping gene, such as -actin and glyceraldehyde-3-phosphate dehydrogenase, as an internal control to normalize the expression of the gene of interest (Giulietti et al., 2001). However, recent studies (Zhong and Simons, 1999; Schmittgen and Zakrajsek, 2000; Goidin et al., 2001; Selvey et al., 2001; Blanquicett et al., 2002; Tricarico et al., 2002) have concluded that it is inappropriate to use these genes as internal controls to normalize gene expression data for a variety of reasons, including the fact that the levels of the control genes are subject to modulation by the experimental conditions and they can vary widely from sample to sample. In fact, according to Bustin (2000), the use of these internal control genes to normalize gene expression should be discouraged. Another approach is to use the levels of rRNA (e.g., 18S rRNA) (Giulietti et al., 2001). However, it has been reported that the expression of 18S rRNA varies significantly between individuals and between biopsy samples taken from the same patient (Tricarico et al., 2002) and that rRNA transcription can be influenced by various factors, including xenobiotics (Spanakis, 1993). Because of the limitations in the use of internal control genes and rRNA to normalize gene expression data, some researchers (e.g., Ogunkolade et al., 2002) have chosen to express their mRNA data as per microgram of total RNA, which can be quantified accurately by a fluorescence-based assay (Jones et al., 1998). However, the use of total RNA to normalize gene expression data does not take into consideration the sample-to-sample differences in the efficiency of reverse transcription (Rodriguez-Antona et al., 2000). In our study (Chang et al., 2003a), we used DNase-treated RNA for the reverse transcription step, quantified total dsDNA concentrations by a sensitive fluorescence-based assay (Singer et al., 1997), and reported the gene expression data as per nanogram of total dsDNA. It should be noted that at the present time, there is not a universal method for normalizing gene expression data generated by real-time PCR (Bustin, 2000). Interested readers should refer to recent reviews on this issue (Bustin, 2000; Giulietti et al., 2001; Bustin, 2002).