Real-time quantitative PCR (QPCR) has made quantifying mRNA transcripts, genotyping, verifying microarray data, and more a simple, automated process. But simple is a deceptive word and when it comes to QPCR, the devil is in the design details.
Primer and probe design, sample preparation, and normalization control selection are some common snares, says Greg Shipley, Director of the Quantitative Genomics Core Laboratory, University of Texas Health Science Center, Houston. While the urge may be to jump right to the cycler (see p. 68), "There's a lot of in silico homework," from determining whether your target has splice variants, to ensuring that it's specific.
The Scientist contacted several QPCR users to identify some of the issues they face and then consulted experts for solutions to their worries.
1. Is my standard really standard?
I-Hsiung Brandon Chen, formerly at the Scripps Research Institute in La Jolla, Calif., was using a Roche LightCycler 2.0 and SYBR Green chemistry to develop an RT-based QPCR assay to confirm gene-expression analysis data collected with DNA microarrays. "In terms of normalization, I think the most common problem I had was to select an optimal reference gene that you use to normalize all genes," says Chen, now at CODA Genomics, Laguna Hills, Calif. "I had to test six to find the one or two that were most stable; the other four did change with treatment." No standard set of internal reference controls exists. How do you find one that's right for you?
Unfortunately, trial and error is standard with internal controls, says Deborah Grove, director of Research Projects at the Huck Institute for Life Science Nucleic Acid Facility, Pennsylvania State University.
Test an aliquot from each treated sample for the level of some presumably stable transcript: for instance, a housekeeping gene such as GAPDH or beta-actin. If it's stable - that is, if the transcript level that QPCR measures is constant across all samples, relative to the amount of input RNA - then you're done; if not, keep trying. Kevin Knudtson, DNA facility director at the University of Iowa, says that while one normalization standard is common, it may not be enough. "If you evaluate at least two targets, you would have a lot more confidence that you were normalizing effectively."
2. Is my RNA degraded?
Ilan Kerman, a research fellow at the Molecular and Behavioral Neuroscience Institute, University of Michigan School of Medicine, also wants to verify microarray-derived gene-expression changes. But Kerman's sample - human postmortem brain mRNA collected from small numbers of laser microdissected cells - doesn't make it easy. Some 20 to 25 hours elapse between death and tissue harvest, says Kerman, an interval that can lead to RNA degradation. "We know it's more degraded than RNA [obtained] from a rat that you sacrifice right away." Furthermore, he says, the changes he's trying to observe range from 20% to perhaps 90%, "definitely less than twofold."
According to James Willey, professor of medicine and pathology at the University of Toledo, Ohio, Kerman faces two potential problems: intersample variation in RNA degradation, and low RNA quantity. "With respect to RNA quality, QPCR is much more forgiving than other gene-expression measurement methods," Willey writes in an E-mail, "especially if the design results in small-size PCR amplicons (e.g., less than 150 bp long)." Moreover, different genes degrade at different rates, he adds.
Willey suggests checking the quality of the RNA on a gel. "If both ribosomal bands are apparent in electrophoretically separated whole RNA and/or if the RNA integrity number (RIN) is high (as described by Agilent), it is likely to yield reliable QPCR results. However, if neither ribosomal band is apparent and/or the RIN value is low, it is likely that a gene-specific criterion for sample quality will need to be established."
With respect to RNA quantity, he continues, one issue is that with very low template amounts, stochastic sampling errors can cause big swings in cycle threshold (Ct) values. "There are many potential causes of a high cycle threshold (e.g., of 34 or 35 or greater)," says Willey, including both low template amounts and gene-specific inhibitors. The way to distinguish the two, he says, "is to include a known number of molecules of an internal standard in each assay."
3. How do I detect low-copy-number templates?
Jill Petrisko, postdoctoral fellow at the University of Idaho, Idaho Falls, uses TaqMan chemistry and a Cepheid SmartCycler II to quantify fungal infection of wheat seedlings. It took months to get her assay working properly. "One of the problems I've had is that I was using a single-copy gene for the assay. It was really hard to detect the fungal DNA in the sample within the assay's linear range," she says. Between 20 and 30 cycles. Her original solution - concentrating the DNA to increase the amount in the reaction - solved her detection problem, but lowered her reproducibility. She eventually switched her primers to detect a multicopy target instead.
"With low-copy-number samples, I try to find ways to increase copy numbers," says Knudtson, either by lengthening the reverse transcriptase reaction (for mRNA quantification) or by increasing the reaction volume. "Many of these kits are sold to be run in 50-µl reactions," Knudtson explains. To cut costs many investigators run reactions at 20 or 25 µl instead, "but if you were going to stay at 50 or even go to 100 µl, you could add more template." An added bonus, says Knudtson: Larger volumes typically give better reproducibility.
4. When is a negative really negative?
Margaret Green, a molecular biologist with the Canadian Food Inspection Agency, Sidney, British Columbia, uses a Roche LightCycler 1.5 with both TaqMan and SYBR Green chemistries for DNA detection via QPCR. "I'm working on developing methods of detection for various targets," Green says. Working in a diagnostic laboratory, it's important to distinguish samples positive for low copy number from actual negatives. "I'd like to know, when a sample has a late Ct value [for instance, greater than 37], can you call that negative?" she asks.
"For me, a Ct of 37 or above is not detectable because there is less than one template molecule available for detection," Shipley writes in an E-mail. A single copy of a 100-bp amplicon can be detected in 35 or 36 cycles, he says, assuming the assay has a near perfect slope that can be transformed into near 100% PCR efficiency. (As the slope increases in value, so will the Ct values, he cautions). "Keeping this in mind, looking for positive results below these values is pointless." Grove suggests upping the template in the reaction, as well as running a nontemplate control reaction.
5. What do I do about inhibitors?
Malcolm Shields, associate professor of microbiology at Idaho State University, uses QPCR to detect and enumerate bacterial species from environmental samples. "We're preparing to do some QPCR on crude samples - sewage," says Shields, who uses an MJ Research Chromo4 system and SYBR Green chemistry. The problem, Shields notes, is inhibitors: Potential sources of inhibition include humic and fulvic acids - carbohydrate derivatives that polymerize and copurify with nucleic acids - plus proteases, detergents, and the like. His solution: Make the DNA as pure as possible, and use spike-in controls to highlight inhibition.
Pamela Scott Adams, director of the Molecular Biology Core Facility at the Trudeau Institute, Saranac Lake, NY, says there's one dead giveaway that you have inhibition. "If you run a standard curve [and] you have 100% efficiency in the assay, then the slope is -3.3," says Adams. If the slope indicates you have greater than 100% efficiency, that's a clear sign that you have inhibition.
"There are some very good commercial columns out there to remove inhibitors," says Willey. "When we do analyses on human primary tissues, we typically use a [phenol-based] TRI Reagent extraction [from Molecular Research Center, Cincinnati] followed by a column [for instance, from Qiagen]. For blood, we have found that the PAX tube is good." Also available, come January 2007, is an Excel system called PREQXCEL, from Jack Gallup at Iowa State University. The system helps design reactions to minimize inhibition (Biol Proced Online, 8:87-152, 2006).
The Quantitative PCR Listserv: A popular Yahoo! Groups listserv, it is frequented by QPCR pros and novices alike. http://tech.groups.yahoo.com/group/qpcrlistserver
Gene-Quantification.info: A compendium of all things QPCR, this site includes information on strategies, normalization, instruments, and more. www.gene-quantification.info
A-Z of Quantitative PCR, edited by S. Bustin: "The 'bible' of real-time PCR," according to Adams, this book "gives a comprehensive review of the basic technology, the various forms of the technology and the theory behind it." www.tinyurl.com/y67dzr
Real-time PCR, edited by M. Tevfik Dorak: Aimed at advanced users and core facilities managers, this book features chapters (including ones by Adams and Shipley) on quantification, normalization, data analysis, primer design, and more. www.dorak.info/genetics/realtime.html
RTPrimerDB: This is a public database of QPCR primers for more 2,200 genes. http://medgen.ugent.be/rtprimerdb
The Association of Biomedical Resource Facilities Nucleic Acids Research Group: Here you can find current and past surveys of QPCR practices, including priming strategies, validation, and more. www.abrf.org/index.cfm/group.show/NucleicAcids.32.htm