Twenty candidate reference genes that showed little variation but high expression in PC12 cells differentiated with NGF and GDNF were first selected from microarray datasets using two independent statistical algorithms. Together with two well studied HKGs, the expression stabilities of these candidate reference genes were further analyzed using RT-qPCR in cells differentiated with other stimuli. From these studies, unexpectedly, RPL19 and RPL29 but not the HKGs, were identified as suitable reference genes that can be used for normalization of gene expression in neuronal differentiation of PC12 induced by a variety of chemical stimuli.
Neuronal differentiation is a process where cells undergo enormous morphological changes, over a period of several days. It is accompanied by substantial biochemical changes including cell cycle exit , changes in metabolism [37, 38] and alteration in structural proteins [39, 40]. Since the commonly used reference genes are mostly structural proteins or enzymes involved in metabolism, it is especially important to validate the stabilities of these genes during the process of differentiation. Many of these studies investigated gene expression changes in PC12 but few have evaluated the suitability of HKGs as normalizers in this model. Our microarray analysis revealed that a group of novel candidate genes was more stably expressed than commonly used HKGs ACTB and GAPDH, suggesting that ACTB and GAPDH may not be ideal reference genes in neuronal differentiation of PC12 cells.
In an effort to gain an insight into the temporal regulation of genes during neuronal differentiation, it is necessary that the reference genes used are stably expressed over a period of days. GAPDH and ACTB have been used for normalization in more than 90% of previous reports , often without proper validation of their stabilities. Numerous publications have reported that such HKGs can be differentially expressed under various experimental paradigms and are therefore inappropriate for normalization [9, 15, 16]. However, there are also recent reports that these HKGs are stably expressed and can serve as reference genes [17, 18, 42]. Most genes, including GAPDH and ACTB, examined in this study were stable at early stages of differentiation (0.5 h or 6 h). However, as differentiation proceeded with dramatic morphological changes and concomitant biochemical changes, the instability of expressions of GAPDH and many of the genes examined was obvious. In the case of GAPDH, this instability issue correlated well with the temporal increase of expression level, which peaked at 28 h and was sustained over a period of 72 h. While GAPDH may still serve as a reference gene for PC12 cells under specific conditions, the validity of using this gene and other less stable ones should be experimentally verified. However, the two RP genes (RPL19 and RPL29) that showed good stability in expression over the period of differentiation provided an optimal pair of reference genes for the entire period of and various experimental conditions for neuronal differentiation.
Among the twenty candidate genes selected, thirteen were ribosomal protein genes, suggesting that the family of ribosomal protein genes may become yet another source of reference genes. Several recent publications have validated and recommended the use of ribosomal protein genes as reference genes [23, 25], while others have reported their tissue-dependent variations . A plausible explanation for such disparity is the large number of ribosomal protein genes present in mammalian systems (80 genes in human, mouse and rat genome), which may be stably or differentially expressed depending on the tissue type or experimental conditions. At present, relatively little is known about these mammalian ribosomal proteins, as compared to their bacterial and archael counterparts . While bacterial ribosomal protein genes exist largely in clusters, the mammalian RP genes are dispersed throughout the genome . Some have suggested that all of these proteins are intimately involved in ribosome production and could be co-regulated. Depletion of a particular ribosomal protein would generally cause a reduction of all other ribosomal proteins in the same ribosome sub-unit . Other reports have shown that some ribosomal protein genes could be regulated independent of others . Recently, extra-ribosomal functions of some of these proteins have been reported [48–51], suggesting that they may be individually regulated. A previous study comparing random ESTs from naïve and NGF-treated PC12 cells, reported an NGF-promoted decrease in the expressions of RPL19 . However, this decrease in RPL19 was not observed in other studies using SAGE  or microarray . Similar to the latter studies, we too did not observe changes in RPL19 transcripts with NGF-treated PC12. Moreover, the SAGE study but not the microarray analysis reported a significant decrease in RPL29 expression. Using both microarray and RT-qPCR, we have also shown that RPL29 was unchanged when the cells were differentiated. The reasons for these discrepancies are unclear and may be due to the differences in methods used. We have shown here by quantitative real-time PCR that some ribosomal protein genes, RPL19 and RPL29, are highly stably expressed and are thus suitable reference genes, whereas others like RPL9 and RPL18 can vary significantly during differentiation.
Unlike some studies that attempted to identify ideal reference genes through meta-analysis of many publically available microarray data which includes a diverse range of tissue types and experimental conditions, this study was designed to specifically identify a set of suitable reference genes for PC12 cells undergoing neuronal differentiation. We have performed both the microarray analysis and RT-qPCR validation on biological samples prepared with the same techniques and reagents, thus minimizing variations introduced by differences in sample preparation methods and assay platforms. We have also systematically evaluated the effect of the use of NFs of inappropriate reference gene(s) on the expression changes of the target genes and the erroneous results they resulted in. Thus, with neuronal differentiation of PC12 cells, scaling with the geometric means of the expressions of RPL19 and RPL29 is recommended for the accurate normalization of gene expression. Whether these two genes are suitable for normalization of neuronal differentiation in other systems remains to be evaluated.