Human placentae were obtained with the approval of the University of Pittsburgh institutional review board (IRB), after uncomplicated elective caesarean deliveries from healthy mothers. Amniotic epithelial (AE) cells were enzymatically dissociated with 0.05% trypsin containing 0.53 mM EDTA-4Na as described previously (8). Viability of the AE cells was determined by exclusion of trypan blue dye and counted with a hemocytometer. Five million AE cells were plated on 100 mm diameter cell culture dishes in culture media with or without 10 ng/ml epidermal growth factor (EGF; PeproTech, Rocky Hill, NJ) for 7 days. The cell culture media was Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine, 1% non-essential amino acid, 55 mM 2-mercaptoethanol, 1 mM sodium pyruvate, 1% antibiotic-antimycotic (All from Invitrogen/Gibco, Grand Island, NY).

Total RNA was isolated from the cell samples at day 0, and day 7 with or without EGF with an mRNA isolation kit (QIAGEN, Valencia, CA). The RNA concentration of each sample was fluorometrically measured (Qubit, Invitrogen). The integrity of RNA samples was further confirmed with the 28S/18S ratio (>1.0) on 1% agarose gel using Gel Logic 100 Imaging System (Kodak Molecular Imaging Systems, New Haven, CT). Two micrograms of total RNA were used to synthesize cDNA (Promega, Madison, WI). The reaction mixture was incubated at 42°C for 60 minutes, followed by heat inactivation of the enzyme at 95°C for 5 minutes. After cooling on ice for 5 minutes, the cDNA was stored at −20°C until analysis.

The TaqMan^® Express Human Endogenous Control Plate (Applied Biosystems, Foster City, CA) that contains 32 genes in triplicate were used to evaluate commonly used reference genes (Table 1). The cDNA samples (5 ng) were subjected to quantitative real-time PCR (qPCR) using an ABI PRISM 7,000 (Applied Biosystems) according to the instructions provided by the manufacturer. As experimental target genes, three additional stem cell specific genes were chosen for analysis. The pre-designed TaqMan probe and primer sets for OCT4 (Hs03005111_g1), NANOG (Hs02387400_g1), and KLF4 (Hs00358836_m1), were selected from the library of TaqMan Gene Expression Assays (Applied Biosystems).

The relative variations of multiple reference genes were calculated by statistical algorithms using the geNorm software package (9). In the first analysis the software returns the internal control gene stability measure (M) for all candidate genes. Furthermore, the software plots two types of graphs. In one of these graphs it plots the average expression stability values of remaining reference genes during stepwise exclusion of the least stable reference gene. This process is repeated until two genes remain. In a second graph the minimal number of reference genes for normalization was determined. This determination is based on the stepwise inclusion of reference genes in the calculation of pairwise variation (V) of a normalization factor (NF). A large variation indicates the added gene has a significant effect and should be included.

The relative expression software tool (REST) was used to quantify mRNA expression of each target gene (10). The REST analysis provides statistical information suitable for comparing groups of treated versus untreated samples by using simple statistical randomization tests with the consideration of multiple reference genes. In this study, gene expression of the day 0 (starting point) samples were used as control group and the samples of EGF (+) and EGF (−) of day 7 samples were statistically compared. Detailed information and free downloading for these statistical software, geNorm and REST, are available at the following websites:

To identify the most stably expressed reference genes in samples from the starting point and two different culture conditions, the expression of a total of 32 genes (Table 1) was measured by quantitative RT-PCR. Data (Ct values) were further analyzed using the geNorm software. The first output showed that M values of all 32 genes were less than 1.5, which is considered as a cut-off value by the geNorm software to use a gene as reference. The average expression stability output showed that there were some reference genes that led to a great variation (Fig. 1A). The exclusion of genes like GUSB or EIF2B1 led to a dramatical decrease in the average. From this analysis resulted that the most unstable reference gene was GUSB (11) and the most stable genes were PPIA (12) and POLR2A (13). The optimal number of reference genes was assessed calculating the pairwise variation (V) between the normalization factor (NF) with a given number of reference genes and the NF value obtained after the inclusion of the following less stable genes (Fig. 1B). The V values were plotted to determine where the inclusion of the additional gene has little effect on the normalization. As seen in Figure 1B, the inclusion of the first three genes as opposed to two was 0.007. According to Vandesompele, et. al. (9), the suggested cut-off value was 0.15 and at least 3 genes must be used as reference. Therefore, in this experimental condition more than three genes were not necessary to perform the normalization. In this present study, we investigated the influence of three reference genes on the interpretation of expression data. It is important to note that the geNorm software determination (M<1.5) has to be considered as the initial screening. Further evaluation is necessary to confirm the valid reference genes.

Expression profiles of the three stem cell specific genes, OCT4, NANOG, and KLF4, were statistically analyzed to investigate the influence of reference gene selection. A total of 9 reference genes were selected and divided into the following three groups based on their expression stability; Group 1: most unstable reference genes, GUSB, EIF2B1, TFRC (Fig. 2A), Group 2: most stable reference genes, PPIA, POLR2A, POP4 (Fig. 2B), and Group 3: commonly used reference genes, GAPDH, 18S, β-actin (Fig. 2C). Each stem cell specific gene expression of EGF(+) and EGF(−) culture conditions was plotted as relative expression to the gene expression of the starting material (day 0). Although the profile of the graph looks similar, the statistical differences were completely different depending on the selection of the reference genes. With the most unstable reference gene, GUSB, the result indicates OCT4 expression was significantly increased in both culture conditions (Fig. 2A). However, when the considerably stable reference genes were used to analyze the data, the OCT4 expression was not significantly increased whereas the KLF4 expression was significantly decreased (Fig. 2B). The results based on the stable 3 reference genes were identical and the interpretation was confident. On the other hand, using some commonly used reference genes and less stable gene group, the statistical data analysis became variable and different from the data obtained with stable genes (Fig. 2A, 2C). When GAPDH and 18S were used as reference genes, the data indicates that OCT4 and KLF4 gene expressions were significantly changed. The significance of OCT4 gene expression in EGF treated groups was not detected when the most popular reference gene, β-actin, was used. The statistical results indicate that the selection of reference genes critically influences the gene expression analysis interpretation.