Widespread interest in gene quantitation and high-throughput assays are putting quantitative PCR back in the spotlight.

More than a decade ago, the polymerase chain reaction (PCR) (1, 2) revolutionized molecular biology by making lab work faster and easier. These days, many researchers expect quantitative PCR to imitate its predecessor's success. Used to determine the concentration of specific nucleic acids, quantitative PCR has been embraced by researchers and clinicians for gene quantitation, pathogen detection, and even process validation in pharmaceutical production. New methods are being published, new products are being developed, and new companies specializing in applying the technique are being formed.

Despite recent attention focused on this technique, quantitation is an old idea--almost as old as PCR itself. "Quantitative PCR has been happening all along," says François Ferré, who heads the gene quantification company Althea Technologies. In the early 1990s, for example, Michael Piatak, Ferré, and others used quantitative PCR to show that HIV viral loads--the degree of infection--in patients' blood were higher than previously thought (3, 4). "Even back in 1989, at meetings focused on PCR, quantitation was a hot topic," Ferré says.

Why all the attention now? There are two main reasons, Ferré and others say. The first is the current emphasis on genomics research. As genes are located, their functions need to be determined, and studies of gene expression become the focus. "The next dimension of research is to figure out what is expressed and how much and when," says Mike Lucero, product marketing manager for PCR at Perkin-Elmer. "That is just as basic as knowing what the DNA sequence is."

The other reason is the emergence of kinetic PCR, also called real-time PCR. In kinetic PCR, measurements are recorded as the reaction occurs. In contrast, the standard endpoint analyses are done using electrophoretic gels after PCR is complete. The basic real-time technique dates back to 1993, but Ferré says it really took off in 1995-1996, when fluorescent techniques that monitor the accumulation of product emerged.

The need for quantitation

But why go to all that trouble? After all, PCR should be inherently quantitative. The number of template molecules should double with each cycle of heating and cooling, resulting in exponential growth (5). To calculate how much product will accumulate, all you should need to know is how much material you started with and how many reaction cycles were run. You should be able to work backward, too, calculating the initial number of copies by knowing the amount of end product and the number of reaction cycles.

That's the theory. In practice, the process is not that straightforward. Amplification eventually reaches a plateau, and it may not be exponential to begin with because of variations in reaction conditions or the presence of inhibitors. Either way, estimates of the number of copies can be wildly inaccurate.

Competitive PCR

The earliest quantitative method was to create standard curves by stopping reactions at various points and determining the amounts of product. The process required radioactive labeling for reasonable sensitivity, and it was tedious, often involving repeated runs at various dilutions of the template. So researchers cheered when a technique called quantitative competitive PCR came along in 1990 (6).

In competitive PCR, two templates are used in a single reaction. One template is the test sample, the initial quantity of which is unknown. The second is the competitor--an internal control that may have nearly the same sequence. In some cases, the competitor differs by only one base pair (bp), so that it includes a restriction enzyme site that the test template does not. Or the competitor may be slightly different in size, causing the two products to migrate different distances on a gel. In any case, the final amounts of both products are compared; and because the initial amount of the competitor is known, the initial amount of the test template can be determined.

Competitive PCR is still widely used, especially for RNA work, but it requires several assumptions that do not always hold true, Ferré says. The target and the standard may not be amplified at the same rate, especially if they have very different sequences. Even a difference of 1 bp can lead to a substantial difference in amplification efficiency, he says. In reverse-transcriptase PCR (RT-PCR), which begins with RNA instead of DNA, the efficiencies of reverse transcription for the template and the standard also might be different. Finally, if restriction enzymes are used, the analysis can be skewed unless all of the competitor is cleaved.

More importantly, individual strands of competitor and test DNA may join together, or hybridize, to form a third DNA species called a heterodimer. This problem, Ferré says, "can seriously complicate the quantitative analysis". Ferré favors the use of external controls, but he's not opposed to competitors. "They can work if you manipulate the system very, very carefully," he says.

Another drawback to competitive PCR is low throughput, says Carl Wittwer, a University of Utah researcher who developed a fluorescent instrument for quantitative PCR. "With competitive PCR, you get your best answer when the concentration of your competitor is near the concentration of your target," he says. Because you don't know that concentration, you must run several reactions using different dilutions of target or competitor--the equivalent of creating a standard curve.

Most competitive PCR techniques also require additional analysis--often by gel electrophoresis--after amplification, says Wittwer. To avoid such extra steps, some researchers are turning to real-time techniques, where quantitation is done without opening the reaction tube. This approach is faster and easier, and it lowers the risk of contaminating the reaction with stray DNA--a single molecule of which could ruin an experiment.

Early kinetic PCR

When many researchers say "quantitative PCR", they mean kinetic or real-time PCR. The technique became well known in 1995-1996, and, like the original PCR, it was an idea waiting to be discovered.

The foundation had been laid years before by Russell Higuchi and colleagues at Roche Molecular Systems. In 1992, they developed a quantitative technique based on a well-known property of EtBr: when EtBr is bound to DNA and excited by UV light, it fluoresces (7). That made EtBr perfect for visualizing bands on electrophoresis gels--something that had been done for years. Higuchi and colleagues realized they could leave the products in the reaction tubes instead and measure the EtBr fluorescence. A year later, the researchers showed that they could continuously monitor EtBr accumulation to obtain kinetic measurements (8). The technique was simple and homogeneous, and it did not require a competitor.

It took about two more years for kinetic-PCR instruments to hit the market. The first instrument was Perkin-Elmer's ABI Prism 7700 Sequence Detection System, followed by Idaho Technology's LightCycler (now manufactured and sold by Roche Molecular Biochemicals). Both instruments use fluorogenic chemistry instead of EtBr, and both are "basically thermal cyclers with optical detection systems," says Lucero. Given that thermal cyclers, optical detection, and fluorogenic chemistry had all been around for years, why wasn't real-time quantitative PCR possible earlier? "It was," Lucero says. "It's just that no one thought of it."

Exonuclease probes

There are two major approaches to kinetic PCR. The first is the Perkin-Elmer method, which combines Higuichi's fundamental kinetic technique with exonuclease probes--originally developed in 1991 by Pamela Holland and colleagues at Cetus (now Roche) (9). The probes capitalize on the ability of some DNA polymerases to cleave unpaired nucleotides from DNA--this is called exonuclease activity. Because the Thermus aquaticus (Taq) polymerase used in PCR has this ability, exonuclease probes are also called TaqMan probes.

In Holland's assay, a single-stranded probe, complementary to a sequence in the target DNA, was used in addition to the two PCR primers. The reaction would begin as usual, with the primers binding to the target and the polymerase extending the second strand by adding nucleotides to the end. But if the probe matched the target, it would also bind. And if the polymerase ran into the probe during the extension process, the polymerase would treat it as an obstacle and remove it by cleaving it into pieces. Because the probe had been labeled with ³²P, the radioactive piece could be isolated on a gel, allowing the researchers to determine how much of the desired product had been produced.

To get away from radioactivity and make the assay more homogeneous, the Perkin-Elmer researchers used fluorescent dyes (10). They placed a fluorescent donor at the 5 end (the upstream end) of the probe. In the middle of the probe, they placed a fluorescent quencher, which kept the donor from giving off light. (These days, the quencher is at the 3, or downstream, end.) If the probe matched the target, the donor was cleaved--just as Holland's radioactive labels had been cleaved--separating the donor from the quencher and increasing the fluorescence (see Figure 1). If the probe did not match, no cleavage occurred. The donor and quencher remained together, and there was no increase in donor fluorescence.

Figure 1. Exonuclease probe.

"We had a quantitative system and a prototype instrument then," says Linda Lee, the Perkin-Elmer researcher who developed the donor-quencher probe. "The next step was to use the system to discriminate between two different alleles [variants of a single gene]." To do that, Lee "multiplexed". She used two donor-quencher probes, one that matched the gene's normal sequence and one that matched a mutated sequence. The two probes used different fluorescent dyes, emitting at different wavelengths, which allowed the signals to be distinguished, and a spectrophotometer measured the fluorescence in real time directly from the reaction tubes.

Hybridization probes

At about the same time, Wittwer and colleagues were developing a different approach to real-time PCR with the LightCycler. The instrument was based on an earlier machine, the RapidCycler, which ran ordinary PCR reactions in glass capillary tubes. The capillaries had been chosen to speed up heating and cooling, which shortened the total reaction time. But Wittwer also saw the capillaries as "an invitation to shine some light" at the reactions. In direct analogy to flow cytometry, which passes labeled cells by a laser, the researchers "essentially passed capillaries containing labeled reactions by a light source," he says.

The researches also developed new fluorescent probes, called hybridization probes (11). Like exonuclease probes, hybridization probes are used in addition to the PCR primers, increasing the specificity, and they can be used to discriminate among different alleles of genes (12). But unlike exonuclease probes, where the signal increases when the fluorescent donor is separated from the quencher, hybridization probes bring two different labels together to allow resonance energy transfer. This happens when both probes bind, or hybridize, to the template DNA. When either probe fails to bind because of mismatches, the resonance energy transfer does not occur, and the fluorescence is much less intense. Thus, the fluorescence is directly correlated to probe hybridization; with exonuclease probes, on the other hand, fluorescence is correlated with probe hydrolysis.

Still, the practical use for the exonuclease and hybridization methods is the same. A fluorescent signal is generated during the PCR reaction and monitored in real time. After about 20 cycles, the signal rises above the background, Wittwer says, and during the next 3-8 log-linear cycles--before the reaction reaches a plateau--the increase is tracked. By analyzing the shape of the curve or by determining when the signal rises above threshold, it is possible to determine the initial template levels. This is a different approach than most researchers are used to, Wittwer says, because what matters is not how much material you end up with, but how quickly you accumulated it.

Variations on the theme

Exonuclease and hybridization methods for real-time PCR are well established, but there are other options, including hairpin probes and hairpin primers, which are named for the way they fold back on themselves. Hairpin probes--also known as molecular beacons--start with their end sequences bound together while the rest of the strand remains unbound, creating a lollipop-like structure called a stem and loop (see Figure 2). While folded this way, the fluorescent reporter at one end of the probe and the quencher at the other end are next to each other in the stem, and there is no fluorescence (13). Binding to single-stranded target DNA straightens the hairpin, moving the quencher's end away and allowing the reporter to fluoresce. Hairpin primers--sold as Amplifluor primers by Intergen (formerly Sunrise primers by Oncor)--are similar, but their fluorescence is generated as they become incorporated into the double-stranded PCR product during amplification, so they do not offer the extra specificty of the probes.

Figure 2. Molecular beacon.

The drawback to all of these fluorescent probes is that they are expensive--about $400 per probe. Thriftier users prefer intercalating dyes, such as EtBr and SYBR Green I, which bind not to the single-stranded template but to the double-stranded product. Dyes are limited in that they do not add extra specificity to the reaction, but Wittwer says researchers don't always need more specificity. "And if you don't need it, don't buy sequence-specific probes," he says. "Any of these real-time techniques will work."

Wittwer adds that most of the methods produce some kind of artifact in the fluorescent signal during the late reaction cycles. The signal becomes nonlinear with SYBR Green I, he says. With exonuclease probes, the fluorescence may continue to rise after the amplification plateau. And with hybridization probes, the signal may decrease artificially during late cycles. According to Wittwer, only hairpin primers appear to be free of artifacts. Nonetheless, he calls this "more an admission of a limitation" than a problem because what matters for real-time quantitative PCR is knowing when the fluorescence appears, which happens long before the artifacts appear.

New detection methods

Other recent alternatives have been based on different detection methods. One technique, developed by Rudolf Rigler and colleagues at the Karolinska Institute (Sweden), is based on fluorescence correlation spectroscopy (FCS) (14). This method uses two PCR primers tagged with different fluorescence dyes. Because FCS can pick out cross-correlated signals resulting from two tags, only the product containing both primers is detected. Rigler says there is no need for an internal or external standard because the cross-correlation amplitude is (inversely) proportional to the number of DNA molecules detected.

Another advantage of FCS is that it allows analysis of a very small volume element, which can reduce the total volume of the reaction, Rigler says. And within that volume, FCS can spot a single molecule, resulting in the detection of as few as 10-25 molecules of initial template. Wittwer says the normal limit of the real-time instruments is ~100 copies of initial template, but fewer than 10 copies can be detected if the systems are fine-tuned. Rigler and colleagues have not tried real-time quantitation yet, but he says it is possible. Their endpoint method has already been turned into a prototype instrument by the Carl Zeiss company.

Another new endpoint assay, developed by Theodore Christopoulos and colleagues at the University of Windsor (Canada), takes advantage of the high sensitivity of chemiluminescence (15). After PCR is complete, the DNA products are attached to the wells of a microwell plate, denatured, and hybridized to specific probes. One chemiluminescent reaction is specific for the test template, and another is specific for the internal control, so it is possible to simultaneously quantitate two different DNA sequences. While acknowledging that chemiluminescence cannot be done in real time because substrate has to be added, Christopoulos says real-time analysis is not always needed, especially when it offers sensitivity of nanomoles-per-liter while chemiluminescence offers picomoles-per-liter. Consequently, he adds, real-time PCR requires more cycles to accumulate a detectable product.

This kind of chemiluminescent endpoint assay also provides an option for users who cannot afford special probes or expensive equipment, he says. "We may end up with a situation where places such as hospitals get fully automated, real-time instruments," he says. "But in research labs and other small setups, people may use a technique like ours, which is simpler."

New challenges

Quantitative PCR already faces competition from other methods. Viral load testing is being done by using branched DNA (bDNA), a chemiluminescent technique from Chiron, and nucleic acid sequence-based amplification (NASBA), an RNA amplification method from Organon Teknika. In branched DNA assays, a tree-like DNA structure binds to the target DNA or RNA, and then labeled probes bind to the tree's "branches". There is no amplification of the target. Instead, the binding of the probes amplifies the signal, indicating that the target has been found. In contrast, NASBA amplifies the target RNA. (The technique is close enough to PCR to be the subject of a lawsuit, but one difference is that PCR, when used for RNA work, does not create new copies of the original RNA, while NASBA does [16]. Another difference is NASBA does not require heating and cooling.) Researchers have already combined NASBA with hairpin probes in a real-time homogeneous assay (17).

Microchips are another emerging competitor for real-time PCR. Because chips can hold large arrays of probes, they are better suited to screening one DNA sample for thousands of possible sequences. "Once you have 10,000 sequences on a chip, it's a lot easier to do one hybridization experiment than to do 10,000 PCR reactions," Wittwer says. "But PCR has an advantage in sensitivity and speed right now."

To keep its place at the forefront, quantitative PCR has to meet several challenges. The first is to increase throughput, Ferré says. "We are in the era of throughput," he says. "The only buzzword around is, 'What is your assay throughput?'" Speeding up sample preparation is part of the solution, Lucero says. The other part of the solution, Ferré says, is to increase the number of reactions an instrument can run. Right now, the ABI 7700 system runs 96 reactions at a time, while the LightCycler runs 32 at a time, although Wittwer adds that the LightCycler's throughput is comparable because the reactions run 3-5 times faster.

More multiplexing is another goal. With more multiplexing, more reactions can be run in the same tube and more direct comparisons can be made. Lee says she has already submitted a paper describing a six-color technique, and Wittwer has developed multiplexing that uses different melting temperatures as well as various colors (12). "If you use color and temperature together, you multiply the multiplexing potential," he says. "That lets you do a lot more with one experiment."

Better sensitivity would also help, Wittwer says. The sooner a real-time fluorescence signal can be detected, the less extrapolation error there will be, so reducing the background fluorescence of the probes would be useful, he says.

Another possible direction is to make real-time PCR "truly real-time", Wittwer says. In the LightCycler, for example, only one fluorescence value per cycle is currently used for quantitation, which means the system sits idle most of the time, he says. It is possible, however, to monitor the fluorescence many times during a cycle to reveal how quickly the probe and target hybridize, he says. That can provide insight into the target's sequence and help determine how much was initially present, because the hybridization speed depends on the target and probe concentrations and how well the sequences match. The information could also be used to adjust the temperature during cycling--a factor that is critical for sensitivity and specificity in PCR. "There's a wealth of information to mine," he says, "and it might as well be processed and analyzed while amplification is occurring."

All of this could add up to the disappearance of earlier techniques, Ferré says. "Real-time PCR will continue to grow because it provides the researcher with a number of highly desirable features, such as an expanded range of quantitation--up to 7 orders of magnitude--and the elimination of post-PCR sample processing that greatly reduces the contamination risk," he says. "It is much more compelling."

But real-time PCR does challenge users to think of PCR differently. People may need to adjust to the new techniques, Ferré concedes, but once they do, they will never go back to the old way of doing things. And if that happens, real-time PCR will truly be a chip off the old block.

References

 (1) Saiki, R.; Scharf, S.; Faloona, F.; Mullis, K.; Horn, G.; Erlich, H.; Arnheim, N. Science 1985, 230, 1350-54.

 (2) Mullis, K.; Faloona, F.; Scharf, S.; Saiki, R.; Erlich, H. Cold Spring Harb. Symp. Quant. Biol. 1986, 51, 263-73.

 (3) Piatak, M.; Saag, M.; Yang, L.; Clark, S.; Kappes, J.; Luk, K.-C.; Hahn, B.; Shaw, G.; Lifson, J. Science 1993, 259, 1749-54.

 (4) Ferré, F.; Marchese, A.; Pezzoli, P.; Griffin, S.; Buxton, E.; Boyer, V. The Polymerase Chain Reaction; Birkhauser: Boston, MA, 1994; pp 67-88.

 (5) Mullis, K. Sci. Am. 1990, 262, 56-61, 64, 65.

 (6) Gilliland, G.; Perrin, S.; Blanchard, K.; Bunn, H. Proc. Natl. Acad. Sci. 1990, 87, 2725-29.

 (7) Higuchi, R.; Dollinger, G.; Walsh, P.; Griffith, R. Bio/Technology 1992, 10, 413-17.

 (8) Higuchi, R.; Fockler, C.; Dollinger, G.; Watson, R. Bio/Technology 1993, 11, 1026-30.

 (9) Holland, P.; Abramson, R.; Watson, R.; Gelfand, D. Proc. Natl. Acad. Sci. 1991, 88, 7276-80.

(10) Lee, L.; Connell, C.; Bloch, W. Nucleic Acids Res. 1993, 21, 3761-66.

(11) Wittwer, C.; Herrmann, M.; Moss, A.; Rasmussen, R. BioTechniques 1997, 22, 130-38.

(12) Bernard, P.; Ajiaka, R.; Kushner, J.; Wittwer, C. Am. J. Pathol. 1998, 153, 1055-61.

(13) Kramer, R.; Tyagi, S. Nat. Biotechnol. 1996, 14, 303-08.

(14) Rigler, R.; Földes-Papp, Z.; Meyer-Almes, F.; Sammet, C.; Völcker, M.; Schnetz, A. J. Biotechnol. 1998, 63, 97-109.

(15) Verhaegen, M.; Christopoulos, T. Anal. Chem. 1998, 70, 4120-25.

(16) Compton, J. Nature 1991, 350, 91-92.

(17) Leone, G.; van Schijndel, H.; van Gemen, B.; Kramer, F.; Schoen, C. Nucleic Acids Res. 1998, 26, 2150-55.