Assessment of preanalytical and analytical variables
Preanalytical variables like the collection and storage of samples and the process of RNA isolation determine the quality of RNA that is used for subsequent quantitative expression analyses. RNA quality, as a summary parameter, characterizes possible negative effects with regard to deficient procedures of collection and preparation of tissue samples. Therefore, the isolated RNA samples from the 25 matched malignant and non-malignant samples were characterized with regard to their concentration, purity, and integrity. Only RNA samples with high quality should be included in this study to avoid erroneous conclusions. However, all RNA samples isolated from renal tissue specimens preserved in RNAlater solution exhibited a high quality. The mean A260/280 ratio of RNA samples was 2.05 ± 0.027 (range from 1.99 to 2.12.) and reflected pure and protein-free RNA. The RNA integrity as an essential quality criterion was characterized by the so-called RNA integrity number (RIN) measured on the Agilent 2100 Bioanalyzer. The mean RIN value (± SD) of all RNA samples was 8.7 ± 0.77 (range from 7.0 to 10.0). The matched malignant and non-malignant tissue samples reached mean RIN values (and SD) of 9.09 ± 0.57 and 8.32 ± 0.75, respectively, whose difference was significant (paired Student's t test; P
There was a strong correlation between the RIN values of the paired samples (rs = 0.635, P s = -0.012 to 0.325; P = 0.956 to 0.113) as it was shown as proof for intact, high quality RNA [27].
Pooled cDNA samples were used as precision control materials for each gene-specific PCR run. These control materials were adjusted to the ranges of Cp values that were characteristic for the particular genes in the tissue samples. Intra-run and between-run analytical performances of the RT-PCR measurements were determined using these control materials. The intra-run precision (n = 12) was 0.26% and 3.56% for ACTB amplifications with a mean Cp-value of 21.26 and the corresponding mean concentration of 4.59 arbitrary units, respectively. The between-run variations (n = 5) ranged from 0.17% (Cp mean 25.63 ± 0.04 of SDHA control cDNA) as the lowest to 0.38% (Cp mean 26.46 ± 0.10 of HMBS control cDNA) as the highest value. Corresponding coefficients of variation for the concentrations data ranged from 2.26% to 10.9%.
Expression levels of candidate reference genes
Paired malignant and non-malignant samples were always measured in the same analytical run to exclude between-run variations. Expression levels of the 10 investigated candidate reference genes (Table 2) shown in terms of Cp-values are given as box-and-whisker-plots in Figure 1. Twenty-five matched samples were included in this study. The boxes represent the median Cp-values and interquartile ranges, the whiskers indicate the 10–90 percentile ranges. Since the expressions measured in the two sample groups were not always normally distributed (D'Agostino & Pearson omnibus normality test), the non-parametric Wilcoxon signed rank test was applied for all significance calculations. Genes with higher expression levels show a lower Cp-value under the specific PCR conditions and lower expressed genes have reversely higher cycle numbers. The Cp-values of all measured genes were between 19 and 31. The genes ACTB, GAPDH, PPIA, RPLPO, SDHA, and TUBB achieved mean Cp-values between 20 and 25 cycles while the lower expressed genes ALAS1, HMBS, HPRT1, and TBP achieved Cp-values between 25 and 29 cycles. As shown in the Figure 1, all genes except PPIA and TBP showed significantly different Cp-values between the malignant and non-malignant samples. PPIA was the highest expressed gene with mean (± SD) Cp-values of 20.04 ± 0.97 in non-malignant and 20.04 ± 0.75 in malignant samples. The TBP gene was the lowest expressed gene with mean (± SD) Cp-values of 29.55 ± 0.53 in non-malignant and 29.72 ± 0.67 in malignant samples.
Significance calculations between gene expression data of paired malignant and non-malignant samples were performed on the basis of concentrations instead of Cp-values, taking into account the different PCR efficiencies as indicated in Table 2. To facilitate the survey of the results, the differential expression of all 10 genes investigated is separately shown for all matched samples (Figure 2a–k). Non-regulated (Figure 2a–c), up-regulated (Figure 2d–g), and down-regulated (Figure 2h–k) genes were grouped. Significantly different gene expressions between malignant and non-malignant samples were observed for all genes with the exception of the genes PPIA (P = 0.605) and TBP (P = 0.371) (Figure 2a–c). Since the smallest mean Cp difference is 0.4 and since retrospective power calculations have yielded a high power (approx. 85%), a type II error can be ruled out with a high probability. The expressions of ACTB (P = 0.0132), GAPDH (P = 0.0002), TUBB (P = 0.0002), and RPLPO (P = 0.0007) were significantly increased in malignant compared with the matched non-malignant samples (Figure 2d–g). In contrast, the levels for ALAS1 (P HPRT1 (P HMBS (P = 0.0002), and SDHA (P = 0.0043) were significantly decreased in malignant tissue parts (Figure 2h–k). Thus, a statistical type I error for these eight genes is very unlikely because of the low P values. In consequence, only the two genes PPIA and TBP fulfil the essential criterion of a reference gene for gene profiling studies in renal cell carcinoma.
Except for TUBB and GAPDH the expression rates of the genes significantly correlated between the malignant and non-malignant samples (rs of 0.429 to 0.790; P = 0.032 to P s = -0.001 to 0.282; P = 0.996 to 0.172), sex (Mann-Whitney test; P = 0.096 to 1.00), and tumour stage (Mann-Whitney test of pT1+2 vs., pT3 or pT1 vs. pT2+3, and pT1 vs. pT3; P = 0.104 to 0.976). The effect of grading could not be estimated, since 23 of the 25 malignant samples were grade 2 and only one had grades 1 and 3.
Expression stability testing of candidate reference genes
PPIA and TBP as the two suitable reference genes were included in the software program geNorm [4,28] to calculate a common normalization factor using the data of these two genes. The application of the normalization factor of these two genes is shown in Figure 2c; it yielded comparable data if only one of the two reference genes PPIA or TBP was used for normalization. We also included all 10 candidate reference genes in the program geNorm to calculate the average expression stability value M for all genes without excluding the genes differently expressed in the matched pairs. The ranking order from the most unstable to the most stable genes was as follows: TUBB, RPLPO, GAPDH, ACTB, SDHA, PPIA, TBP, HMBS, ALAS1, and HPRT1. The last remaining two genes ALAS1 and HPRT1 achieved a stability value of 0.426 as the most stable genes. It is remarkable that all genes had average expression stability M values less than 1.5 that is defined in the program as stability cutoff value. However, these results show that the program was not able to identify the genes with significant expression differences between malignant and non-malignant sample pairs.
Using the NormFinder program [29,30] as another free tool available on the internet to validate the expression stability of the candidate reference genes, the two genes PPIA and TBP also achieved the best stability values (Table 3). With the aid of this program, PPIA was identified as the most stable single gene with a stability value of 0.074. Thus, the stability data calculated with that program as a combined estimate of intra- and intergroup expression variations of the genes studied reflect to a certain extent the expression differences of the genes observed in the matched pairs.
Expression levels of target gene influenced by normalization genes
The expression levels of the target gene ADAM9 is used here as an example to demonstrate the effect of different normalization genes on the relative gene expression data. We determined the ADAM9 mRNA expression in the same 25 RNA sample pairs as used for the reference gene search. The ADAM9 expressions were normalized using different strategies (Figure 3). ADAM9 expression values were related either to the single non-regulated reference genes PPIA and TBP, the mean ratio of both genes (mr), the normalization factor (NF) obtained for these two genes using the program geNorm, or to the three up-regulated genes GAPDH, ACTB, and TUBB as well as to the two down-regulated genes ALAS1 and HPRT1. To better compare the effect of these different normalization approaches, the relative ADAM9 mRNA expression in non-malignant samples was set 1.0 and expression rates in the matched malignant samples were calculated as multiples (Figure 3). A significant up-regulation of ADAM9 mRNA in malignant tissue samples was proved when the normalization approach was performed either with the two stable reference genes PPIA and TBP or with the mean ratios of both genes and the normalization factor calculated for the two genes using the geNorm program. In contrast, the normalization of ADAM9 expression data on the up-regulated genes resulted in a partly decreased ADAM9 mRNA expression, whereas the normalization procedure with the down-regulated genes resulted in an about three-fold enhancement of the up-regulation of ADAM9 mRNA expression in malignant tissue samples.
Answers to all your Biology Questions
