Quantitative Real-Time PCR Is Not More Sensitive than “Conventional” PCR

Molecular methods, essentially based upon PCR, have become an indispensable tool in the diagnosis of infectious diseases. Real-time quantitative PCR (qrtPCR), as a novel technology, has revolutionized molecular diagnostics by adding reliability and speed (15, 28). However, Apfalter et al. (1) have stressed the need for caution in analyzing results from both qrtPCR and “conventional” PCR (cnPCR, herein designing any form of endpoint detection). We are also concerned by the fact that many reports tend to generalize the idea that qrtPCR in general has an improved sensitivity compared with cnPCR or enzyme-linked immunosorbent assay (ELISA)-PCR. It is absolutely undeniable that qrtPCR has greatly improved our molecular diagnostic practices and, as such, represents a considerable progress in molecular diagnosis. Indeed, it has many objective advantages over cnPCR, particularly speed, broad dynamic range of target DNA quantitation, and reduction of contamination. But it should be stressed that qrtPCR is not necessarily more sensitive than cnPCR. (Sensitivity in this paper is taken in its broader sense, whether “analytical,” “absolute,” or “diagnostic,” and may be assimilated to the lower limit of detection of the assay. Obviously, when compared examples are given, we considered studies using the same type of sensitivity [essentially “diagnostic”].)

This relates to a more general consideration: in many reports of comparative studies between a given qrtPCR assay and a cnPCR assay, the conclusions drawn by the authors seem to apply to qrtPCR in general, as if their assay represented “the” qrtPCR (per se) and the other represented cnPCR per se. Multiple examples of this can be found in the literature, particularly in abstracts or introduction sections (e.g., references 12, 13, 18, 27, 30-33, 37, and 38). The same phenomenon occurred when microbiological diagnosis by cnPCR was becoming more widely used in diagnostics in the 1990s and when “PCR” was considered as a single highly sensitive method, regardless of the assay parameters and the primers and probes used and whatever the performance characteristics. It is our intention to remind readers that in any such study the findings hold true only for the particular (PCR) assay used but certainly not for all PCR assays of the same type (whether nested PCR, cnPCR, or qrtPCR, etc.). What we emphasize here essentially concerns “in-house” or laboratory-developed PCR assays, which still constitute the vast majority of PCR diagnostic assays in microbiology (1). Our examples will focus the discussion on parasitology, mycology, and bacteriology, as these are our fields of competence and because commercial tests are less frequently available for the detection of these microorganisms than for virus detection; however, this commentary also concerns virology.

Considerable interlaboratory variations can easily be shown by examining reports from different groups for exactly the same diagnosis (1, 2) or from comparative multilaboratory studies or external quality assessment reports (5, 10, 20, 26, 34). What is the cause of these variations?

Although this issue has already been extensively reviewed (1), we believe it is extremely important for the understanding of the biologist and clinician, and hence for patient care, to underline the following: (i) the PCR is not just one technique but a method encompassing a number of techniques, each with a variable outcome depending on a variety of factors; and (ii) qrtPCR is essentially a sophisticated technological advance of PCR, but it has many of the same limitations with respect to interlaboratory variations. A simple comparison can be made here with serology. One would never consider claiming that all ELISAs were more sensitive than all immunofluorescence assays simply because of reports that an ELISA performed better than an immunofluorescence assay for the detection of a particular antibody. For reasons that are difficult to understand, this is often not the case with PCR. Comparisons between laboratory-developed assays, qrtPCR or otherwise, should likely be made with even greater caution.

It may be useful to recall that, for a given microorganism in a given biological sample, the sensitivity of a PCR assay primarily depends upon three factors: the physicochemical conditions of the reaction, the concentration and nature of the (microorganism) DNA target, and the selected PCR primers and probes.

Ideally, the reaction conditions should be submitted to a large number of subtle cross-modifications that constitute the so-called “optimization” of the PCR conditions (essentially the hybridization temperature for primer annealing, and the Mg²⁺Cl₂, DNA Taq polymerase, and primer concentrations). This process, which may take as long as 3 to 6 months in cnPCR, aims at enhancing both the sensitivity and the specificity of the reaction (9). It is noteworthy that this optimization is a considerable factor of variation among PCR assays, in addition to other factors mentioned. Hence, the use of different commercial preparations of DNA polymerases or other chemical reaction compounds and the time and care spent upon “optimizing” the reaction conditions are variation factors as essential as the DNA target and more so than the primers. Along the same lines, the use of universal master mix preparations, by reducing the possibilities of optimization by the operator, inevitably may influence the assay sensitivity. Also of note is that the total amount of DNA in the template preparation, which includes DNA from the pathogen and its host, varies from sample to sample in routine practice. This, in turn, directly influences the amount of free Mg²⁺ ions in the reaction mix and hence the enzymatic activity of the DNA polymerase. Thus, high concentrations of human DNA can decrease the sensitivity of molecular assays designed to detect infectious pathogens.

The number of copies of the DNA target in the genome of the microorganism is well known as an essential factor that affects sensitivity. Still, the fact that two PCR methods use the same DNA target, or even the same primers, does not necessarily ensure identical results.

A striking illustration of this is provided by assays designed to detect Leishmania donovani, a protozoan parasite and the causative agent of visceral leishmaniasis. The presence of a highly (5,000×) reiterated sequence, the kinetoplast minicircle, in the mitochondrial genome of this organism allows the design of PCR assays that have been termed “ultrasensitive” (22, 23, 35) compared with PCRs that target less repetitive sequences. Nevertheless, among the variety of primers designed for this target, it has been shown that some primer pairs targeting the same minicircle in the same Leishmania species yield a sensitivity 100-fold lower than that of others (22). Thus, an assay using suboptimal PCR primers that target this extremely repetitive target may not reach the sensitivity of a more classical PCR that targets a moderately repetitive gene. Conversely, several PCR assays targeting the >100-fold-reiterated spliced leader RNA gene in Leishmania have consistently shown a mediocre analytical sensitivity compared with those targeting the <20-fold-repeated ribosomal DNA (about 1 and 100 parasites, respectively, detected per reaction) (4, 17, 21, 22, 29). Still, in the majority of cases, the degree of reiteration of the target is proportional to the sensitivity of the assay (i.e., the more copies of the target are present, the easier it is to detect). Hence, a cnPCR targeting the kinetoplast minicircle is more likely to be more sensitive than a qrtPCR targeting a less repetitive sequence; but exceptions exist (N. Bourgeois et al., unpublished data).

Although qrtPCR may be seen as less open to “optimization,” and hence to variation, the same considerations do apply, with the same consequences. Thus, for Toxoplasma gondii, another protozoan parasite, qrtPCR assays using exactly the same (B1 gene) target and different primers have been reported with analytical sensitivities varying by a 200-fold factor, i.e., from 0.05 to 10 genome equivalents per reaction tube (6, 7, 11, 13, 14, 19, 25, 33, 36). Still, such variations are due not only to the choice of the primers but also, at least as importantly, to the optimization of the method: indeed, identical primers may yield widely different performances in different laboratories (P. Bastien and the Molecular Biology Group of the French National Reference Center for Toxoplasmosis, unpublished data). Also, it is noteworthy that the same authors may find up to 1,000-fold variations in qrtPCR depending on whether they use SYBR green I or fluorescence energy transfer hybridization probe detection (32), which reiterates that qrtPCR is not a single method. Strikingly, many of these detection thresholds are much higher than those obtained using cnPCR assays on the same DNA target: for example, a 0.5 genome equivalent was easily detected by Chabbert et al. (9) using cnPCR, a result far from being necessarily achieved by qrtPCR assays designed for the same pathogen. Furthermore, some reports specifically describe the low sensitivity of qrtPCR compared with cnPCR, whether for Toxoplasma (18) or for other microorganisms (16, 38). Although this commentary is written from the view of bacteriology or parasitology and mycology, all of the mentioned aspects are holding true in the field of virology. For example, cnPCR (PCR-enzyme immunoassay) was also found more sensitive than qrtPCR in cytomegalovirus detection (24). Consequently, it cannot be assumed that qrtPCR per se is more sensitive than cnPCR: this should be considered as a misleading statement.

Another factor that influences the robustness of the in vitro amplification process, especially in diagnostic medical microbiology, is the total reaction volume, more particularly the proportion between the volume of template DNA preparation input and that of the PCR master mix solution. When applying higher reaction volumes, traces of inhibitory substances potentially present in the template DNA preparation are often diluted to a subcritical level. When comparing qrtPCR and cnPCR for the detection of Mycobacterium tuberculosis in respiratory specimens, the observed analytical sensitivity of the assay appeared proportional to the volume of template DNA input in the PCR (8). Since most of the modern qrtPCR devices are designed for total reaction volumes of 50 μl or less, the restricted template DNA input volume may limit the analytical sensitivity of the corresponding assay.

Additional factors influencing PCR sensitivity include the nucleic acid extraction method and the genetic variability of the DNA target. There is increasing evidence that differences in the efficiencies of the extraction methods may have a significant influence on assay sensitivity (40). Moreover, the “optimization” process should result in a perfect match between the nucleic acid extract (and therefore the extraction method) and the PCR conditions. Finally, using different input volumes and different methods of sample preparation with clinical specimens (enzymatic pretreatment followed by ethanol precipitation, spin columns, silica-coated magnetic particles, etc.) may result in different recovery rates of the target organism genome as well as in different concentrations of human genomic DNA in the eluate, both factors being able to play a role in the assay performances (39). Thus, even an optimized PCR assay may yield poor results when DNA has been extracted by a suboptimal method for that assay (P. Bastien, unpublished data). As regards genetic variability, when selecting the DNA target and primers, one should ensure that the corresponding sequences are not subject to variation. Such variations may lead to sequence mismatches and to a suboptimal annealing between the primers and the (mutated) target. Such sequence variations may also introduce more-stable secondary structures into the amplicons which may unexpectedly hamper the amplification process and reduce the analytical sensitivity. Finally, the effect of primer quality and the sometimes striking variations between the determined and reported oligonucleotide concentrations from primer synthesis companies are frequently underestimated.

All these effects are cumulative, and it is obvious that their influence on the PCR process (both sensitivity and specificity) is more dramatic in low reaction volumes typically applied in modern qrtPCR assay formats.

Finally, before following the fashionable trend of switching from cnPCR to qrtPCR, one should carefully consider the pros and cons of the discussed amplification formats. For example, the physics used for amplicon detection are similar in both assay formats and, with identical PCR volumes, the use of SYBR green or TaqMan or fluorescent resonance energy transfer probes does not necessarily generate a higher intensity of fluorescence than “classical” intercalating dyes in agarose gel analysis. When colorimetric amplicon detection strategies (such as line blotting and enzyme-linked oligosorbent assay, etc.) are implemented in the analytical workflow of a cnPCR assay, the level of analytical sensitivity achieved may be up to 10 times higher than that achieved with the use of fluorescent probes in qrtPCR (U. Reischl, unpublished data).

We hope that the reason for this commentary is not misunderstood. This is certainly not a “rear-guard struggle” for cnPCR in an era of qrtPCR. There is no doubt that qrtPCR represents significant progress for all molecular diagnostics in microbiology, and it has been rapidly replacing cnPCR in the vast majority of its applications. This is due to a number of undisputable technical advantages of qrtPCR, such as speed, robustness, reproducibility, and low contamination risk, in addition to the information provided by quantification and melting curve analysis of amplification products. Simply, one must be aware that, strictly speaking, the analytical performances of qrtPCR are not “automatically” better than those of cnPCR. The use of a qrtPCR assay alone does not guarantee excellence. The advent of any technique in biology is accompanied by variants, themselves playing upon variables that inevitably influence the test performances. As stressed by several others (1, 9, 20), the essential factor in setting up a novel biological (and particularly molecular) diagnostic test is a constant questioning of the optimization of the reaction conditions, as well as a minimum degree of proficiency. In this respect, a recently published roundtable of experts (3) stressed that “caution should be exercised in introducing qrtPCR into laboratories with little molecular experience and no R&D backup; this requires staff with technical skill and awareness to develop and validate these assays and to support any technical difficulties experienced during their routine performance.” With regard to publications, in addition to the critical recommendations made by Apfalter et al. (1), we would generally recommend that phrases such as “qrtPCR proved more sensitive than cnPCR” be completed with the terms “for this particular assay” as well as by a precise mention of the methods used by the authors. This obviously also remains valid for comparisons between two qrtPCR assays.