to BioTechniques free email alert service to receive content updates.
How Reliable is Real-Time PCR? | PCR Feature

05/02/2012
Sarah C.P. Williams

Two people can perform the same real-time PCR experiment and get different results. Researchers are quantifying this variability to understand its source and how to fix it.

Bookmark and Share

When a cell turns up the production of a gene, the first measurable effect is the increase in messenger RNA, the intermediate molecule between a gene and its eventual protein product. For scientists who want to follow cells as they change—based on health, environment, or internal cues—measuring the change in mRNA molecules is the best way to track how a cell’s genes are turned up, down, on, or off.

Last year, Vaillancourt started a series of such RT-qPCR experiments on a new Bio-Rad real-time PCR machine. But when two of her students did the same experiment, they got different results. Source: Vaillancourt





“There is already a lot of work regarding the standardization of qPCR methods,” says microbiologist Nuno Cerca of the University of Minho in Portugal. “However, I do feel that it’s underappreciated. Despite normalization guidelines, we can still have results that are reproducible, but not reliable.” Source: Nuno Cerca

“The MIQE guidelines have now been cited over 600 times in the peer-reviewed literature, with the paper now being the eighth most-cited Clinical Chemistry publication,” says Bustin. “So, the message has started to get through to the research community.” Source: Stephen Bustin

Since the mid-1990s, reverse-transcription quantitative PCR (RT-qPCR) has been the go-to way to track the concentration of mRNA molecules in a cell. But each step of RT-qPCR can introduce error into the experiment. Now, a handful of scientists are taking a step back to see how they can improve their data by optimizing how they apply RT-qPCR to their areas of research.

“There is already a lot of work regarding the standardization of qPCR methods,” says microbiologist Nuno Cerca of the University of Minho in Portugal. “However, I do feel that it’s underappreciated. Despite normalization guidelines, we can still have results that are reproducible, but not reliable.”

The existing guidelines that Cerca is alluding to are the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines. These guidelines were first published in the journal Clinical Chemistry (1) in 2009 by a team of researchers led by Stephen Bustin, a molecular biologist at the Queen Mary University of London. The guidelines suggest details that researchers should incorporate into their qPCR experiments to ensure accurate results. They fall into four key areas: sample quality control, assay design, PCR efficiency, and normalization.

“The MIQE guidelines have now been cited over 600 times in the peer-reviewed literature, with the paper now being the eighth most-cited Clinical Chemistry publication,” says Bustin. “So, the message has started to get through to the research community.”

But other researchers continue to perform qPCR and publish their results without going through the optimization steps recommended by MIQE. “I think it’s a mix of people both not being aware of the guidelines, and just not wanting to follow them for convenience’s sake,” says Cathy Vaillancourt of the Institut Armand-Frappier Research Centre (INRS) in Canada.

Back to Basics

In her lab at the INRS, Vaillancourt studies how the placenta is different between healthy pregnant women and those with gestational diabetes or pre-eclampsia. To understand the conditions, she and her colleagues look at the varying levels of gene expression in different placentas. They rely on RT-qPCR to quantify how much mRNA is being made from any given gene of interest.

Last year, Vaillancourt started a series of such RT-qPCR experiments on a new Bio-Rad real-time PCR machine. But when two of her students did the same experiment, they got different results. So, she set out to understand what introduces variability into RT-qPCR experiments and what researchers can do to fix it. She quickly zoomed in on reference genes as her team’s major source of variability.

To control for the fact that different samples might start out with different concentrations of total RNA, researchers rely on reference genes—genes that are thought to be expressed at the same levels in all cells. If a reference gene has twice as high levels in one tissue, based on an RT-qPCR experiment, then it means the researcher must halve the concentration of any gene of interest in order to truly compare relative values.

“The placenta has a lot of RNA-degrading enzymes,” says Vaillancourt, “and a lot of mRNA can be degraded before the samples have been isolated.” This, and the fact that diseases like diabetes and pre-eclampsia could cause such widespread genetic changes that even reference gene levels could be changed, made Vaillancourt begin to think that their choice of reference genes was introducing large amounts of error into her team’s RT-qPCR experiments.

Vaillancourt went back to the basics; instead of taking it for granted that the commonly used reference genes would be best for her work, she used two computer programs to analyze the stability of the reference genes and work out which were the most stable and therefore had the most consistent levels between her healthy and diseased placenta samples (2). It’s a step recommended by MIQE but not one that Vaillancort had performed before.

What she found was beyond her expectations. The levels of the genes varied so much between samples that if she chose one reference gene versus another, it would totally reverse the trend seen in a separate gene of interest.

“Using one reference gene would make this gene of interest appear to be increasing in expression,” she says. “While another reference gene could make it seem to be decreasing in expression.”

The lesson her team took home is the importance of reference gene selection in qPCR experiments. “Now, we’ll be doing this every time we do an experiment,” says Vaillancourt. “We’ll always try to find the best reference genes for that particular experiment—the genes that have the least variability between the tissues we’re studying.”

Kit vs. Kit

At Cerca’s lab at the University of Minho in Portugal, the question wasn’t one of reference genes, but of which RNA extraction kit worked best for his research. Cerca studies biofilms, collections of bacteria that grow along a solid surface—rather than being free-floating in a liquid media. While Cerca had found it easy to extract RNA from E. coli cultures, when he started trying to isolate RNA from biofilms of other types of organisms, he began having problems. Some RNA extraction kits didn’t seem to work that well.

“There was already some literature that described that different RNA kits can result in different RNA quality,” says Cerca. “But, different quality of RNA does not necessarily mean that we would have different gene expression outputs. And this is, ultimately, the most important question: what kit would give me reliable, meaningful results at the lowest price possible?”

Cerca’s team set out to compare five commercially available RNA extraction kits on three different foodborne biofilm-forming bacteria (3). They began with identical samples in each kit, and ran RT-qPCR experiments to quantify the RNA from each extraction kit. While the RNA yield and purity from the kits had no effect on the RT-qPCR experiments, the RNA integrity—how much is degraded—had an effect on the experimental results, and varied between kits.

“I have to admit that I was surprised with some of our results: in some cases we had three-fold less RNA concentration but over 10-fold differential gene expression,” says Cerca.

His team found a kit that works best for them—one optimized for biofilms. But it won’t be the best for everyone. Going through the comparison exercise is key. And identifying—for each researcher—what aspects of their qPCR experiments need optimization is important.

Bustin, who spearheaded the MIQE guidelines, is now moving on to more comprehensive education for researchers who want to learn real-time PCR—or who want to improve their results. He’s developed a series of iBooks that walk scientists through the steps to optimize qPCR experiments.

“A well-designed assay will not be too different from another well-designed assay for the same target,” says Bustin. “But two badly designed assays will give different results.”

References

  1. Bustin, S. A., V. Benes, J. A. Garson, J. Hellemans, J. Huggett, M. Kubista, R. Mueller, T. Nolan, M. W. Pfaffl, G. L. Shipley, J. Vandesompele, and C. T. Wittwer. 2009. The MIQE guidelines: Minimum information for publication of quantitative Real-Time PCR experiments. Clinical Chemistry 55(4):611-622.
  2. Lanoix, D., A.-A. A. Lacasse, J. St-Pierre, S. C. Taylor, M. Ethier-Chiasson, J. Lafond, and C. Vaillancourt. 2012. Quantitative PCR pitfalls: The case of the human placenta. Molecular biotechnology (April).
  3. Fran├ža, A., J. C. Bento, and N. Cerca. 2012. Variability of RNA quality extracted from biofilms of foodborne pathogens using different kits impacts mRNA quantification by qPCR. Current microbiology (April).

Keywords:  PCR qPCR RT-qPCR