Validated normalization is crucial to obtain reproducible qPCR data of genes of interest. In this context, normalizing to internal reference genes has become the most popular method to control for experimental errors introduced by the multitude of steps in this analysis. Several studies point out that the expression of reference genes may vary under different experimental conditions[13–17]. This implies the necessity of validating these genes in each new experimental setup.
To our knowledge, this is the first study that describes the stability of 18S rRNA, ActB, GusB, Arbp, Tbp, CycA and Rpl13A in the DG of rats one week after FS. Both, geNorm and Normfinder were used to rank the analyzed reference genes by their expression stability. This rank order differed slightly between both methods, probably because both tools are based on different mathematical models. Other studies have also described similar ranking discrepancies between geNorm and Normfinder[18, 24]. Interestingly though, both programs agreed on the three most stably expressed genes, being CycA, Tbp and Rpl13A. These converging results stress the significance of including these genes in the normalization factor. In addition, both programs also agreed on ActB and 18S rRNA as the least stably expressed genes. Comparison of these data with those of recent studies revealed similarities and differences. For instance, Bonefeld et al. validated eight reference genes in rat hippocampal tissue and also identified CycA and Rpl13A as the most stably expressed genes and ActB and 18S rRNA as the least stable genes. Also Pernot et al. found that CycA and Tbp were stably expressed in hippocampus samples from a mouse model of TLE, obtained across different phases of the disease. However, in contrast to our study they also observed a stable ActB expression. This discrepancy emphasizes the importance of validating reference genes in each experimental model.
Accurate normalization requires inclusion of multiple reference genes. Geometric averaging of the most stable reference genes is a validated method to obtain a reliable normalization factor. Based on a cut-off value of 1.5, geNorm indicated that the normalization factor should be based on five reference genes. However, according to the geNorm manual, this cut-off value can be set differently. geNorm calculates the optimal number of reference genes by pairwise variation analysis. In that respect, the trend of changing V values after adding additional genes can be used to obtain an estimate of the number of genes that should be included in the normalization factor. Determination of the optimal number of reference genes is always a trade-off between accuracy and practical considerations, but a minimum of three most stable reference genes is generally recommended. As indicated in Figure 2B, a pairwise variation of 0.198 was observed after adding the third most stable gene. Inclusion of the fourth or fifth most stable gene influenced only slightly the pairwise variation. The high V6/7 is caused by the high average M value of 18S rRNA, indicating that this gene is highly variably expressed under the present experimental conditions. The Acc.SD calculated by Normfinder suggested the use of six reference genes, though the additive value of genes four to six is minimal. Considering the pairwise variation values, the Acc.SD, and practical issues such as the available amount of RNA, we conclude that the geometric mean of CycA, Rpl13A and Tbp should be used to obtain an accurate normalization factor in this experimental setup. If the RNA yield allows the inclusion of an extra reference gene, GusB and Arbp may be added to this panel. These data also show that 18S rRNA is unfit as reference gene in this model.
Inclusion of the geNorm/Normfinder selected reference genes in the normalization factor, revealed an increased Cnr1 expression in animals that experienced FS (FS+). This finding is in agreement with quantitative western blot data from Chen et al.. This upregulated Cnr1 disappeared when expression levels were normalized to ActB and 18S rRNA, underscoring the suggestion that these ‘classical’ reference genes are not suitable for our experimental setup. In line with this observation, several studies have reported that including 18S rRNA or ActB in the normalization factor altered mRNA expression levels compared to normalization to geNorm proposed genes[18, 20]. As a possible explanation for erroneous normalization when 18S rRNA is used as reference gene, it has been suggested that this may relate to an imbalance between messenger RNA and ribosomal RNA.