NATURE BIOTECHNOLOGY VOL 17 SEPTEMBER 1999 http://biotech.nature.com 835

COMMENTARY

Quantitative application of the reverse tran-scription-polymerase chain reaction (RT-PCR) has been the subject of considerabledebate for the last decade1-5. The goal ofmost quantitative RT-PCR methods is touse PCR product yield as a measure of rela-tive differences in mRNA template abun-dance1,6. Because the efficiency of RT is usu-ally assumed to be constant, the quantita-tive capacity of the PCR has been the prima-ry focus of debate. Early on, the feasibilityof quantitative PCR was questioned becauseof two theoretical constraints: (1) Given theexponential nature of the process, initiallysmall tube-to-tube differences in amplifica-tion efficiency would grossly affect the finalyield of PCR products1-3; and (2) PCR prod-uct yield could only provide a valid measureof template input during the exponentialphase of amplification2,4.

The development of competitive PCRmethods allowed investigators to addresssuch theoretical concerns and quantitativePCR was born2,3. In competitive PCR thefinal measure of template abundance issolely dependent on the initial ratio of tar-get to competitor DNA templates. Thus,each reaction is internally controlled andtheoretical arguments regarding the vari-ables of amplification are largely irrele-vant3,5. Given the initial skepticism, thestrict internal controls of competitive PCRwere essential to prove that quantitativePCR was possible. Since its development,however, few investigators have questionedwhether or not competitive PCR methodsare actually necessary to achieve a quantita-tive RT-PCR assay.

Competitive RT-PCR is often impracti-cal for routine applications such as com-paring the gene expression profile (i.e. mul-tiple mRNAs) of large numbers of RNAsamples7. Furthermore, when the primarysource of variability is not the PCR, but theexperimental system itself (e.g. environ-mental6 or animal7 studies), some investi-gators have questioned if precision shouldalways take precedence over practical utili-ty. Thus, alternative, high-throughputassays have been developed for comparingtemplate abundance using conventionalRT-PCR methods6-10. These “semi-quanti-tative” PCR methods are generally consid-

ered inferior, however, to competitive PCRbecause there is no internal control foramplification efficiency1-5. In the originaldescription of competitive PCR, Gillilandet al.3 provided evidence that inclusion of acompetitor DNA was essential for accuratequantitation. Of three more recent side-by-side comparisons, however, competitiveand “semi-quantitative” RT-PCR assayswere found to produce equivalent measuresof template abundance9,11,12.

The assertion that competitor DNAs areessential to quantitative PCR is based on theoriginal tenets that standard PCR amplifica-tion is (1) highly variable, and (2) proceedsto maximal levels in the plateau phaseregardless of differences in template input.Although widely accepted as fact, the empiri-cal evidence does not support either hypoth-esis. Following standard PCR amplification,a 10-20% coefficient of variability in productyield is typically observed from replicateDNA samples8,9,13,14.

Regarding the second point, the results ofHalford et al.14 demonstrate that PCR prod-uct yield in the plateau phase is not simply allor none. After 35 cycles of PCR, maximalamplification of PCR products occurs in allreactions containing greater than ∼ 5 x 104

templates. However, as template becomeslimiting (∼ 5 x 102 - 5 x 104 templates), PCRproduct yield in the plateau phase is depen-dent on the logarithm of template DNAinput14. Although the results are inconsistentwith current PCR theory, Halford et al.14

demonstrate that a major factor has beenoverlooked in past theoretical considera-tions: primer-dimers.

Despite optimal design, primers sponta-neously form amplifiable primer-dimers ata low rate based on the sheer number ofoligonucleotides in the PCR (∼ 200 trillionper ml). Once formed, primer-dimers areefficiently amplified and serve as endoge-nous competitors of the PCR. As templateconcentration becomes limiting, primer-dimers constitute an increasing fraction ofthe total number of PCR products. When areaction approaches the plateau phase,primer-dimers compete for reactants andthus inhibit the amplification of specificPCR products14. This phenomenon

accounts for two related observations: (1)The lower limit of template detection inPCR run to the plateau phase is not the the-oretical value of 1, but is normally300–1000 templates per reaction13,14, and(2) when template is limiting, PCR amplifi-cation reaches the plateau phase before spe-cific PCR products accumulate to maximallevels (fig. 4 of ref. 1, fig. 3 of ref.14).

Ten years ago, the question was posed“What conditions must be satisfied toachieve a quantitative RT-PCR assay?” Basedlargely on theoretical considerations, theanswer was that (1) each reaction had to beinternally controlled by a competitor DNA1-5

(2) the competitor DNA had to be nearlyidentical in sequence to the target DNA3,and (3) product yields had to be measured inthe exponential phase of amplification2,4. Adecade later, none of these conditions haveactually proven to be necessary for quantita-tive RT-PCR5,9,14.

So the question remains “What are theessential prerequisites for establishing aquantitative RT-PCR assay?” To date, a greatdeal of effort has been spent defining the“correct” conditions that allow for quantita-tive RT-PCR. Given the inherent quantita-tive capacity of the method14, however, per-haps it is time for RT-PCR to be treated likeall other quantitative assays. That is, regard-less of the specific PCR conditions chosen,the essential prerequisites for a quantitativeRT-PCR assay should be that (1) a standardcurve demonstrates the range over whichPCR product yield provides a reliable mea-sure of mRNA input, and (2) the number ofsamples tested allows for statistical analysisof differences in PCR product yield.

1. Freeman, W.M., Walker, S.J., and Vrana, K.E.Biotechniques 26, 112-125 (1999).

2. Wang, A. M., Doyle, M. V., and Mark, D. F. Proc. Natl.Acad. Sci. 86, 9717-9721 (1989).

3. Gilliland, G., et al. Proc. Natl. Acad. Sci. USA 87,2725-2729 (1990).

4. Wiesner, R. J., Beinbrech, B., and Rüegg, J. C.Nature 366, 416 (1993).

5. Siebert, P. D. and Larrick, J. W. Nature 359, 557-558(1992).

6. Chandler, D.P. J. Indust. Microbiol. & Biotech. 21,128-140 (1998).

7. Halford, W.P., Gebhardt, B.M., and Carr, D.J.J.Virology 238, 53-63 (1997).

8. Hill, J.M., et al. Anal. Biochem. 235, 44-48 (1996).9. Vandyever, C. and Raus, J. Cell. Mol. Biol. 41, 683-

694 (1995).10. Sykes, P.J., et al. Biotechniques 13, 444-449 (1992).11. Zhang, C., et al. Biochem. Biophys. Res. Comm. 223,

450-455 (1996).12. Haberhausen, G., et al. J. Clin. Microbiol. 36, 628-

633 (1998).13. Martell M et al. J. Clin. Microbiol. 37, 327-332 (1999).14. Halford, W.P., et al. Anal. Biochem. 266, 181-191

(1999).

The essential prerequisites for quantitative RT-PCR

William P. Halford

William P. Halford is in the department ofmicrobiology, University of Pennsylvania, 221Johnson Pavilion, Philadelphia, PA, 19104-6076. Email: halford@mail.med.upenn.edu.

Perhaps it is time for RT-PCR to be treated like allother quantitative assays.

© 1999 Nature America Inc. • http://biotech.nature.com©

199

9 N

atu

re A

mer

ica

Inc.

• h

ttp

://b

iote

ch.n

atu

re.c

om