| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Briefs |
1 Department of Pharmacology and Toxicology, University of Alabama at Birmingham, Birmingham, AL 35294
2 Department of Cancer Biology, Dana-Farber Cancer Institute, Boston, MA 02115
aaddress correspondence to this author at: Department of Pharmacology and Toxicology, University of Alabama at Birmingham, 1824 6th Avenue South, WTI 608B, Birmingham, AL 35294, e-mail Richie.Soong{at}ccc.uab.edu
Because of their low detection limits, PCR and the adapted kinetic (real-time quantitative) PCR have been used extensively for the detection and quantification of nucleic acids (1)(2)(3). However, at the detection limits of an assay, mathematical models (4)(5)(6) and experimental evidence (7)(8)(9)(10)(11)(12) have shown that nucleic acids are detected inconsistently and quantified imprecisely. With this potential unreliability, it is essential that indicators be used to identify the concentration at which analysis is occurring to ensure the accuracy of results. Conventional PCR indicators, however, currently provide only an indirect assessment of the concentration at which analysis occurs. The total nucleic acid content of a sample, commonly used to define the adequacy of sample loading, does not precisely reflect specific gene concentrations. Reference (control or housekeeping) genes, qualitative detection of which is often used to validate assay and sample integrity, are often present at concentrations different from those of the target genes of interest (13). In kinetic PCR, the inclusion of fluorescent probes allows the monitoring of reaction kinetics, which enables the measurement of a crossing point (CP), or cycle threshold, at a partial cycle number at which fluorescence becomes detectable above background signal (14). CP values are linearly proportional to gene copy number in an inverse logarithmic relationship (14)(15), and this correlation is central to algorithms for determining gene quantity (16)(17). In this study, we hypothesized that CP values might be more accurate indicators of the reliability of an analysis than conventional indicators because of their direct relationship with gene concentration. The results of this study provide experimental evidence to support this hypothesis; we also describe models for use of CP values as indicators for assessing the reliability of analysis.
To test our hypothesis, we simulated the analysis of low concentrations of the putative colorectal micrometastasis marker cytokeratin 20 (CK20) (16) and the reference genes porphobilinogen deaminase (PBGD) and ß2-microglobulin (ß2M) and compared the capabilities of conventional indicators (total nucleic acid content and qualitative reference gene detection) and CP values to identify samples with unreliable analysis.
We prepared, by serial dilution, four sample series comprising 10 000, 1000, 100, 10, and 1 plasmid DNA copy for CK20, PBGD, and ß2M (16) and 20 000, 2000, 200, 20, and 2 pg of SK29-Mel-1 RNA (18). Five replicates of each concentration were analyzed for CK20, PBGD, and ß2M by kinetic reverse transcription-PCR on the LightCycler instrument according to the protocol in the LightCycler CK20 Quantification Kit (cat. no. 3118835). A reverse transcription reaction was performed for each SK29-Mel-1 RNA replicate, amounting to 25 reverse transcription reactions. The plasmid DNA and SK29-Mel-1 cDNA series for each gene were analyzed in separate PCR runs, amounting to six individual runs of 25 samples each. Detection mixtures for CK20 and PBGD were obtained from the LightCycler CK20 Quantification Kit, and the detection mixture for ß2M was obtained from the LightCycler ß2M Housekeeping Gene Set (cat. no. 3146081). CP values were determined by the LightCycler Analysis Software (Ver. 3.5), using the data-driven second-derivative maximum function. All reagents and instruments were from Roche Diagnostics unless otherwise specified.
The replicate detection frequency, mean CP values, and their SDs for each gene in each sample are provided in the Data Supplement that accompanies the online version of this Technical Brief (http://www.clinchem.org/content/vol49/issue6/). In accordance with previous findings (7)(8)(9)(10)(11)(12), in samples with high concentrations, gene detection was consistent, with high CP reproducibility and linearity (R2 >0.999) between concentrations and CP values, whereas in those with low concentrations, genes were detected inconsistently or not detected, CP reproducibility was decreased, and linearity was lost.
When we assessed analysis reliability, using conventional indicators, there was little correlation between the occurrence of an unreliable analysis and the sample nucleic acid content or among the three genes. Detection was inconsistent or absent in samples containing 200, 20, or 2 pg of CK20 RNA and samples containing 2 pg of PBGD RNA, whereas ß2M was detected consistently in all samples. With increasing dilution, CP reproducibility markedly decreased for CK20, PBGD, and ß2M in samples containing 2000, 20, and 2 pg of RNA, respectively. Linearity was lost (R2 <0.999) in the same samples in which RNA was inconsistently detected, i.e., samples containing 200 and 2 pg for CK20 and PBGD, respectively. Hence, for this samples series, nucleic acid content or qualitative reference gene detection provided a poor indication of the reliability of detection and quantification. The correlation between the occurrence of unreliable analysis and copy numbers among the genes was better. Inconsistent detection and loss of linearity occurred in samples with single copies of CK20 and PBGD. Reduced CP reproducibility occurred in samples with 10 copies for all three genes, highlighting that the inadequacy of conventional indicators was likely attributable to their lack of direct correlation with specific gene quantities.
Assessing CP values as potential indicators showed that inconsistent detection (Fig. 1A
), poor CP reproducibility (Fig. 1B
), and loss of linearity between CP values and concentration (Fig. 1C
) occurred within specific ranges of CP values. Importantly, these correlations were independent of the gene or the nucleic acid type being analyzed, suggesting that CP values could be useful indicators for reliability of analysis of other genes and samples of unknown concentration.
|
To define models for the use of CP values as indicators, we considered that four zones of PCR reliability indicating consistent detection with high CP reproducibility (white zone), poor CP reproducibility (gray1), inconsistent detection and loss of linearity (gray2), and unlikely detection (black) could be mapped from the coincidence of the different types of PCR reliability with CP values. Using the lowest mean CP value for poor CP reproducibility, we defined the white/gray1 boundary at CP 33.06. We chose the CP value of 37.01, corresponding to the first observation of inconsistent detection, as the gray1/gray2 boundary and the theoretical CP value at which CK20 in 20 pg of RNA should have been detected to indicate the gray2/black boundary. Extrapolating the CP difference of 3.73 between the mean CP values for 20 000 and 2000 pg of CK20 RNA, we determined that detection of CK20 in 20 pg of RNA should have occurred around CP 42.80 [31.61 + (3.73 x 3)]. Thus, using this model, we could determine the reliability of detection or quantification of a test sample by its CP value and its corresponding zone.
Apart from defining the reliability of a result, indicators also identify the adequacy of samples to provide a reliable result, particularly when detection is likely to be inconsistent. To define a model for this, we observed that the difference in CP values between two genes (
CP) remained relatively constant, independent of sample concentration (Fig. 1C
), although this was valid only for reliably analyzed samples. We surmised that addition of a reference gene CP to an expected CP(target - reference) would enable the CP of the target gene, and hence its reliability of analysis from its detection zone, to be estimated. At a CP(CK20 - ß2M) of 11.88, based on the analysis of 20 000 pg of RNA, this model accurately predicted the reliability of CK20 detection for each respective ß2M CP and sample concentration (Table 1
). Using this model, we could thus determine the likely adequacy of a sample to provide reliable analysis for a given target gene from the CP for a reference gene and knowledge of an expected CP(target - reference).
|
For the purposes of brevity and clarity, we have used a selected sample series to demonstrate the concepts of this study. In exploring the wider application of our models, we have observed similar trends with other genes, sample sources, and detection formats (TaqMan probes), using the same assay system (results not shown). However, analyses on other assay systems and instruments have indicated that these systems differ in their detection limits and hence the positioning of the zonal CP boundaries. Nevertheless, the concept of zones of reliability and the capabilities of CP values to indicate these zones remained. These results have suggested a wider application for our models, but they also indicate that with different assay systems, preliminary determination of the boundaries is required. This would require only analysis of a replicated sample series serially diluted to the assay detection limit, similar to that of this study.
An important element of our models is the derivation of an expected CP(target - reference). Our experience indicates that this value can be determined with use of cell lines or pilot samples representative of the samples to be analyzed at a high concentration (results not shown). Alternatively, it could be determined mathematically. In this context, using boundaries defined in this study and assuming ideal product doubling with each PCR cycle, a CP(target - reference) of 4 would make it conditional for samples to have reference gene CP values <29.07 (33.06 - 4) to ensure that a 16-fold (24) difference between two genes is quantified reliably. However, it is also important to recognize that the constancy of CP(target - reference) is dependent on similarities in the PCR efficiencies between two genes. Using methods described previously (16), we determined that the efficiencies of the CK20, PBGD, and ß2M PCRs for RNA samples in the linear range were 93%, 100%, and 98%, respectively. By linear regression analysis, these efficiencies were not significantly different from each other, suggesting that constancy of CP(target - reference) for a given sample type could be assumed. Clearly however, depending on the predictive precision required, it may be prudent to ensure equivalence in efficiency before making this assumption.
In summary, lack of reproducibility among laboratories and lack of adequate guidelines for assay standardization (11)(19)(20)(21) continue to hinder the clinical application of many new assays. One problem has been that, with the improved detection limits and dynamic ranges of assays, there has been little revision of the indicators used (22). This study demonstrates that CP values may provide an assessment of PCR reliability that is superior to conventional indicators, and we devised models for their implementation. Because CP values and CP(target - reference) are already integral elements in kinetic PCR quantification (16)(17), use of these parameters provides direct and quantitative indicators that require no additional preparation. These measures potentially provide for more intuitive selection of samples and interpretation of results, but this promise awaits further validation and application in future studies.
References
CT) method. Methods 2001;25:402-408.[CrossRef][Web of Science][Medline]
[Order article via Infotrieve]
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  M. L. GATTON, J. M. PETERS, K. GRESTY, E. V. FOWLER, N. CHEN, and Q. CHENG DETECTION SENSITIVITY AND QUANTITATION OF PLASMODIUM FALCIPARUM VAR GENE TRANSCRIPTS BY REAL-TIME RT-PCR IN COMPARISON WITH CONVENTIONAL RT-PCR. Am J Trop Med Hyg, August 1, 2006; 75(2): 212 - 218. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 O. Suslov and D. A. Steindler PCR inhibition by reverse transcriptase leads to an overestimation of amplification efficiency Nucleic Acids Res., November 27, 2005; 33(20): e181 - e181. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Zimmermann, A. El-Sheikhah, K. Nicolaides, W. Holzgreve, and S. Hahn Optimized Real-Time Quantitative PCR Measurement of Male Fetal DNA in Maternal Plasma Clin. Chem., September 1, 2005; 51(9): 1598 - 1604. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
