Login

Join for Free!
112223 members
table of contents table of contents

A study demonstrating the suitability of the two housekeeping genes PPIA and …


Biology Articles » Molecular Biology » In search of suitable reference genes for gene expression studies of human renal cell carcinoma by real-time PCR » Discussion

Discussion
- In search of suitable reference genes for gene expression studies of human renal cell carcinoma by real-time PCR

The essential result of our study evaluating 10 candidate reference genes for normalizing gene expressions of clear cell RCC was that only the two genes PPIA and TBP did not differ in their expression in malignant and in non-malignant tissue pairs. Consequently, only these two genes fulfil the criterion of expression stability between matched tissue samples and could be recommended as accurate normalizers for relative gene quantification in clear cell RCC samples. In the following, this conclusion and data will be discussed, taking into account four main aspects of the current study: (a) the particular study conditions concerning the preanalytical and analytical variables, (b) the panel of the candidate reference genes studied, (c) the validation of suitable reference genes, and (d) the limitations of the study.

Preanalytical and analytical study conditions

Like in our previous experiments with reference genes [8,9], the preanalytical and analytical design of our study is defined by three characteristics: 1) use of paired malignant and non-malignant tissue samples from the same nephrectomized organ; 2) use of high-quality RNA samples as precondition for reliable RT-PCR measurements; 3) high analytical performance realized by measurements on the LightCycler. We consider these preconditions essential to achieve a suitable normalization for gene expression studies. Recently, Huggett et al. [17] recommended a similar approach.

The use of paired samples from the same RCC instead of unpaired samples for comparative gene expression studies has the particular advantage of minimizing the interindividual variation effect and increasing the statistical significance of the study [8,9]. The correlations between the corresponding gene expressions in the malignant and non-malignant tissue samples support the view that interindividual expression variations do occur. In consequence, differential gene expressions can be easier ascertained so that the use of paired samples results in a more exact validation regarding the suitability of reference genes than a study using unpaired samples. Therefore, the use of paired samples is the method of choice in identifying suitable reference genes for differential gene studies between malignant and non-malignant samples. In clinical research the detection of different gene expression levels between non-malignant and malignant tissues is an important objective. The results are starting points for further studies on translational stage with the final goal of developing new diagnostic markers or therapeutic approaches. However, in validating of suitable reference genes, one should above all, consider that the expression of candidate reference genes can be influenced by physiological factors like age, sex, pathological factors like the tumour characteristics stage and grade or other biological conditions [23,29,31-34]. The results of our data proved that the expression of all genes was not dependent on age, sex, and tumour stage.

Another characteristic of our study was the use of high-quality RNA samples as an important prerequisite to obtain reliable RT-PCR results [17,35] for selecting stable reference genes. For expression measurements, we used only RNA samples with an A260/A280 ratio > 1.95 and RIN values > 7.0, as an index of high purity and integrity of the RNA samples. We controlled the quality of each RNA sample with the Agilent 2100 Bioanalyzer. The principle of the bioanalyzer technology is based on microcapillary electrophoresis [27,36]. Electropherograms and gel-like images can be visually evaluated and an expert software can generate an RNA integrity number (RIN). The RIN algorithm evaluates six regions of the electropherogram in addition to the conventional ratio of 28S rRNA to 18S rRNA [27,36]. The RIN allowed a user-independent assessment of RNA integrity. Although mRNA only accounts for 2–5% of total RNA, the RIN value provides a relevant information concerning downstream applications [35]. Based on the correlation analysis of RIN values to RT-PCR data, Schroeder et al. [27] defined for their experiments a RIN value of 6 as a threshold for high and low RNA quality while the low quality RNA was considered unacceptable for downstream measurements. Collection, storage, and processing of the tissue samples are the essential preanalytical variables that affect the RNA integrity as a global indicator of all detrimental processes occurring during the time between the collection of samples after surgery and isolation of RNA [27,37]. The isolated high-quality RNA in our experiments and the proof of the non-correlation of RIN values with the expression results confirmed that possible distorting effects by non-intact RNA could be avoided. In addition, the use of RNA samples with similar degrees of integrity is shown in our experiments by the strong correlation of the paired RNA samples (rs = 0.635, P [38] suggested that samples of comparable RNA intactness could be used for microarray analysis despite a certain degradation. Our previous experiments have shown that not only the interval time between the removal of the organ and preservation, but also the method of preservation is decisive for the integrity of RNA. Comparisons of RIN values measured in RNA isolated from paired tissue samples stored in RNAlater solution or snap-frozen in liquid nitrogen immediately after surgery showed significantly higher values using RNAlater-preservation. Therefore, we consider the use of RNAlater solution as method of choice in preservating renal tissue for obtaining intact RNA for reliable expression results. RNAlater was also recently recommended for preserving RNA integrity in routine clinical kidney biopsy material [39].

The analytical performance of the RT-PCR measurements was characterized by low variation coefficients of Cp-values ranging from 0.17 to 0.38% in between-run precision experiments. This high analytical precision, together with the use of paired samples and their measurement in the same analytical run, enabled us to carry out data analysis with high statistical probability.

Panel of the candidate reference genes studied

As briefly outlined in the Introduction, a systematic study concerning the suitability of reference genes for normalization in RCC expression studies has been lacking so far. Schmid et al. [7] performed a Medline search about reference genes used in renal tissue in general. Although, as reported, we included in our PubMed search only publications concerning RCC (Table 1), we, like Schmidt et al. [7], found GAPDH and ACTB to be the most frequently used genes for normalization. It should be mentioned that the search strategy in PubMed using the MESH terms "renal cell carcinoma" in conjunction with the terms "RT-PCR" and "gene expression" was limited, since published studies not indexed with these terms cannot be found. The reviewer drew our attention to this point after our study was finished. For example, the study by Janssens et al. [40] using the gene mitochondrial ATP synthase 6 (mATPsy6) in addition to GAPDH and HMBS (named by the authors as PBGD) as reference genes in various cell lines and tissues was not found with this strategy and was not included in this study. However, we believe that despite this limitation, the overview in Table 1 served as good starting point for our study. The selection of eight out of the 10 candidate reference genes in our study resulted from this search to facilitate comparison, while the two additional genes TUBB and SDHA were included into the panel of candidate reference genes due to their utility shown in other studies [6,8,9]. All these candidate reference genes studied are widely used in other studies. Our study was performed to search for stable housekeeping genes in a large panel of candidate reference genes covering a broad expression range. The two genes 18S-ribosomal RNA and beta-2-microglobulin also commonly applied as normalizers in RCC samples (Table 1) were not included in our study panel because they belong to the highly expressed and regulated genes in renal tissue [6]. Moreover, the use of 18S-ribosomal RNA as reference gene would be only possible if random primed reverse transcription was carried out for cDNA synthesis.

Validation of the suitable reference genes

To identify the suitable reference genes as normalizers in gene expression studies, several strategies including computer programs have been recommended [1,17,23,29,40-43]. There is no doubt that the absence of the differential expression of the candidate reference gene examined in the study groups in question or under conditions to be compared is the strongest proof of suitability. Therefore, we proposed to use a two-step strategy for identifying suitable reference genes according to our study design of matched malignant and non-malignant samples [8,9]. First, the expression of the candidate reference genes between the respective study groups or conditions is compared using the appropriate paired tests. It can be assumed that genes with significantly different expressions are not suited to target gene normalization, since they are affected by the study condition in question. Those genes should be excluded as normalizers. In the current study, the expression of all genes except that of PPIA and TBP varied in the matched malignant and non-malignant samples. The difference in expression was observed not only in the two genes ACTB and GAPDH, which are most frequently used as normalizers in RCC studies, but also in the gene HPRT1. The latter was recently recommended as single reference gene for gene expression studies in cancer research [44]. It was also identified as a suitable reference gene in RCC samples by Haller et al. [23] using the equivalence test. However, this discrepancy to our results might be due to the fact that Haller et al. [23] only studied 10 paired samples with an obviously lower statistical power compared with that in our experiments with 25 paired samples.

In the second step of our approach, we generally calculated the best-performing reference genes using the geNorm and NormFinder programs [8,9]. Whereas in prostate cancer and bladder cancer experiments 13 out of 16 and six out of nine genes, respectively, remained as suitable genes for the further calculations with geNorm and NormFinder [8,9], only the two genes PPIA and TBP could be considered for these analyses in the current study. It is worth mentioning that the geNorm program was unable to detect candidate reference genes that were characterized as unsuitable reference genes in the first step of suitability testing. In contrast, NormFinder recognized PPIA and TBP as the two best-performing reference genes when all candidate reference genes were included for the calculation (Table 3).

A target gene normalization was used as an example to illustrate the validation of a suitable reference gene selected from a panel of candidate reference genes. We used "a disintegrin and metalloproteinase domain 9" (ADAM9) as the target gene. Without going into details about the possible significance of ADAM9 in RCC tumourigenesis, we can say that the various relative quantification methods shown here can result in serious gene quantification errors if unsuitable reference genes are used (Figure 3). We recommend the use of PPIA and TBP for normalizing expression results using the mean ratio of relative quantification with the two reference genes. Although the advantage of using both reference genes for normalizing is not clearly evident in our study, other authors showed a more accurate normalization when more than one reference gene was used [4,7,33].

Limitations of this study

Some limitations of this study should be mentioned. The first limitation could be seen in the limited number of samples used in the current study. In contrast to this limitation, it is remarkable that even with the possibly low statistical power due to the limited number of samples, clearly significant results of eight candidate reference genes were obtained. Thus, as explained in the previous section, the risk of a type II error and also of a type I error is rather unlikely as the problem with small studies does not exist in our study. Second, the present study is limited to the clear cell RCC subtype. However, that type is the most frequent malignant RCC. The expression of the two recommended reference genes PPIA and TBP in the papillary or chromophobe RCC subtypes was not studied and it would be necessary to confirm their potential use in further studies. Third, the study was partly unbalanced with regard to sex, tumour stage, and grade. The study included 21 males, but only 4 females. However, since there were no differences for all genes studied, we can assume, despite this unbalanced design, that the expression did not depend on sex. Similarly, the tumour stage obviously did not influence the expression level as shown in the Results. Twenty-three out of 25 samples had the tumour grade 2. Whereas it can be assumed that at least PPIA and TBP are suitable reference genes for grade 1, their usefulness as normalizer in grade 3 RCC samples remains to be verified. Fourth, although we examined the most comprehensive panel of potential housekeeping genes in comparison with other studies, we measured only 10 genes, excluding genes very rarely used for renal cell carcinoma due to the limited sample material available (Table 1). Thus, the question remains unanswered whether these or other genes including the mitochondrial ATP synthase 6 gene mentioned above are equivalent to or more suitable than the recommended genes PPIA and TBP.


rating: 0.00 from 0 votes | updated on: 31 Aug 2007 | views: 8514 |

Rate article:







excellent!bad…