Introduction

Quantitative reverse transcription PCR (RT-PCR), which detects fluorescence intensity during the amplification process, has greatly evolved and overcome the limitations of classical RT-PCR assays, giving an extraordinarily easy and accurate method to quantify nucleic acids (1,2). Thus quantitative RT-PCR, with a very large dynamic range (seven to eight orders of magnitude) and a high sensitivity (as few as 10 molecules), has been widely utilized in many aspects of biologic research including oncology (3,4,5,6). Nevertheless, the best method to quantify and calculate the absolute or relative expression of target genes by real-time PCR is still under debate. The standard curve method and comparative cycle threshold (C_T) method and their variants have been developed and used for relative expression quantification in different laboratories (7). In the standard curve method, the input amount for unknown samples is calculated from the standard curve of a specific gene and normalized to the input amount of a reference gene, which is also calculated from its standard curve (8). The comparative C_T method detects and calculates relative gene expression using the formula 2^-▵▵C_T (9). This formula is based on the assumptions that amplification efficiencies of the reference and target genes are approximately equal and that the amplification efficiency is close to 2. Nevertheless, these two groups of methods use a common parameter: the value of C_T. C_T value is the fractional number of cycles required to reach a particular threshold fluorescence signal level. It can be decided manually or automatically using different methods and algorithms (10,11). In some apparatuses, such as the LightCycler®, the term crossing point or Cp is used for the same concept. Each amplification has its characteristic C_T value, which is determined by the initial concentration of target gene. Although widely used, these methods are always criticized for their inherent complexity and imprecision. The problem is due to the absence of information concerning exact amplification efficiency. As mentioned by many researchers, the variation of amplification efficiency is a non-negligible source in bias of PCR results (9,12,13,14,15). Another problem for C_T-based methods is that the exact variation of initial numbers of target gene, which is believed to vary between 6%–20% (8,16), cannot be directly quantitated by these methods.

Recently, a new quantification method has been proposed by Liu and Saint (17) and then revised by Rutledge (18). This method, called sigmoidal curve-fitting (SCF), appears to be an attractive one because it gives a simulation of the whole PCR process. The new method obviates the need for construction of relative standard curves and the need for a series of validation experiments that are the prerequisite for the comparative C_T method, providing a more accurate, faster, and more efficient quantification method.

To test this SCF method, we developed a quantitative RT-PCR assay to measure the expression of two genes (XRCC4 and HIF1α) in human tumor samples both with the classic method (standard curve method) and with SCF. These genes were chosen because, in our laboratory, we believed they might be useful for predicting radiosensitivity. The gene product of XRCC4 participates in the repair system of doublestrand DNA breaks (19,20), while HIF1α is a subunit of the transcription factor, hypoxia-induced factor 1 (HIF-1), which is the most reactive gene in response to hypoxia, another source of radioresistance during radiotherapy (21,22,23,24,25,26). In this study, we compared the results obtained with both methods, we proposed an equation for relative quantification based upon the SCF method, and we tested its sensitivity and reproducibility by measuring XRCC4, HIF1α, and the housekeeping gene, HPRT, messenger RNA (mRNA) levels in different human cell lines and tissue samples.

Materials and Methods

Biological and Patient Samples

The biological samples used in this study included cell lines and patient tumor samples. Two cancer cell lines were used to check the reproducibility of the experiment: HT29 cL.19A (human colorectal adenocarcinoma, kindly supplied by M. Laboise, INSERM U239, Paris, France) and HCT116 (human colorectal adenocarcinoma, supplied by J. Bourhis, Villejuif, France). Cells were grown in their recommended medium and controlled for mycoplasma contamination by DNA fluorochrome staining every month during the experiment. For clinical samples, 21 primary non-small cell lung cancer samples were collected (from June 2001 to November 2004) immediately after surgery. Tumor samples were examined by a pathologist and stored at −190°C according to culture collection guidelines. The total RNA of each sample was extracted, and the expression of the target genes and reference genes were measured.