Quantitative, Inexpensive PCR for Research, Diagnostics
dPCR works by partitioning a sample into many individual real-time PCR reactions. After amplification, those reactions containing at least one copy of the target are counted as positive; the remainder are negative. Depending on the number of individual subsamples, software applies a statistical algorithm that generates an absolute count for the number of target molecules in the original sample—all without reference to standards or controls.
The main competing dPCR technologies employ thousands or millions of droplets as the reaction “vessels.” By contrast, partition dPCR uses nanofabricated physical reaction chambers.
Related Article: The Emergence of Digital PCR
In Bio-Rad’s (Hercules, CA) droplet system, each sample is mixed with reagents, probes, and primers and partitioned into 20,000 nanoliter-size droplets, each encapsulated within an oil emulsion. The droplets are then returned to a fresh microwell and undergo temperature cycling in a standard thermal cycler, where they replicate to end point. Fluorescence detection and application of a calculation algorithm provide a precise concentration value for the target sequence in the original sample.
“The advantages [of] droplets compared with physical partitioning [are] flexibility, scalability, and lower cost,” says Viresh Patel, PhD, marketing director at Bio-Rad.
dPCR was conceived in the early 1990s and eventually patented by Cytonix (Beltsville, MD) and the U.S. Department of Health and Human Services. Cytonix is still active in nanotech materials and instruments.
The idea behind dPCR is to dilute and partition samples to the point where, ideally, one or zero target molecules exist in each subsample. Post-PCR, the number of fluorescently positive subsamples provides a direct count of the target sequence’s original concentration.
Bio-Rad’s approach treats the partitioned nanodroplets as a Poisson distribution in which a statistically predetermined number of droplets contain zero, one, two, or more target molecules. After counting fluorescently positive samples, software back-calculates and corrects for those wells containing multiple targets.
“The statistics work out and are very reliable because we have a very large number of droplets, approximately 20,000, per sample,” notes Paula Stonemetz, director, diagnostic business development at Bio-Rad. “Some droplets will have two or more copies, but calculating their numbers is straightforward.”
Because all droplets are cycled to where reagents have run out, droplets originally containing one, two, or more target molecules have very close to the same fluorescence signal. Those lacking a copy show much weaker spurious or background fluorescence.
The nanodroplets and their architecture are the remarkable backstory of droplet dPCR. After addition of reagents, each sample is added to a droplet-generation cartridge with a proprietary oil formulation. As the sample pushes through a microfluidic channel, the perpendicular flow of oil breaks the sample into nanoliter droplets and encapsulates each droplet.
“The surfactant and oil chemistry we developed [are] the real magic behind droplet dPCR,” Stonemetz says. “Once formed, the oil-encapsulated droplets are quite stable. They do not coalesce or merge as long as they are not handled too roughly and [we] use standard pipetting technique.”
Droplets are returned to a fresh microplate, in which they are cycled in a standard thermal cycler. When the PCR exhausts itself, the droplets enter a reader similar to a flow cytometer that reads 1,000 droplets per second.
Upping the Numbers
“What’s exciting here is you can generate 40,000 or even a 100,000 droplets if you like if you need to push sensitivity down to one part per million,” Stonemetz adds. “This is difficult to achieve with physical partitioning because of the sheer numbers.”
Roopom Banerjee, president of dPCR instrument developer RainDance (Billerica, MA), likens dPCR to megapixels on a digital camera; the more you have, the higher the resolution. RainDance’s RainDrop instrument, a platform for both next-generation sequencing and droplet dPCR, splits samples into ten million droplets compared with 20,000 droplets for the Bio-Rad platform and the same number of physical compartments for the Thermo Fisher instrument. RainDrop is also capable of multiplexing—measuring more than one target per test. Most diseases are caused by multiple genetic factors, not single genes.
Why ten million droplets? “If you can’t discriminate between one and ten copies directly, and if both produce a positive, that’s no longer a digital result,” Banerjee says. “You’ve lost the ability to count.”
He provides as an example HIV viral-load monitoring, where the difference between five, ten, or 50 copies in a blood sample reflects either the subtype or the progression of the virus inside the immune system. The number in turn triggers decisions regarding the type or aggressiveness of treatment.
“You should not have to use statistics to obtain what should be a truly digital yes [or] no answer,” Banerjee says. “You want an answer that says one or zero, which allows quantitation of both the high numbers and the low numbers.”
Alicia Burt, senior product manager for digital PCR at Thermo Fisher Scientific (San Francisco, CA), refers to dPCR as the “third wave” of PCR, the first two being endpoint and quantitative PCRs. “dPCR can be used for many of the things we do with other types of PCR, but where it stands above is looking for something extremely rare, the needle in the haystack.” Even qPCR would have difficulty picking out a transcript with a 0.1% occurrence.
Thermo Fisher’s variation of dPCR is chip-based. Samples are diluted and spread onto a single-use substrate containing 20,000 wells—one chip per sample. The benefit of wells is that every reaction chamber is precisely the same size, which provides continuity between and within experiments. Dilution is such that on average 1.6 copies of the target exists in each well. Some, of course, have zero, some one, two, etc.
As with droplet PCR, each well undergoes endpoint PCR before reading, with both relative and absolute quantitation possible. Based on a Poisson statistical treatment similar to the one used by Bio-Rad’s instrument, the software reports back how many copies of the target existed in the original sample, although wells containing one copy give the same binary positive signal as those containing several copies. According to Burt, having reaction wells of identical size reduces the likelihood that the calculations will misrepresent the initial concentration.
Thermo Fisher has improved on the Poisson statistical treatment to account for effective well volumes. “Each well has a capacity of 755 picoliters. The new Poisson-Plus algorithm goes an additional step by correcting for any variation in the volume of liquid actually present in the wells. “It allows us to correct for exactly what’s happening on this chip,” Burt tells Lab Manager. “Poisson-Plus is a better predictor of original concentrations, particularly at higher sample concentrations.”
dPCR fits exquisitely in the assemblage of complementary molecular biology techniques for diagnostic and prognostic medicine. Burt explains how it fits in. “You do sequencing when you’re looking for something but don’t know what it is yet.” For example, a solid tumor sample that has not been characterized for genotype. “dPCR comes in once you know what you’re dealing with, say, one or more known mutations, because it can find that needle in a haystack. If you’re monitoring samples over time, dPCR is easier, faster, cheaper, and in most cases more sensitive than sequencing.”
Burt mentions a recent American Cancer Society meeting that featured many breakout sessions on liquid biopsies, which are genetic tests on blood instead of solid tissues. Compared with a solid biopsy, liquid biopsies via dPCR (or some other gene-profiling technique) are relatively noninvasive. dPCR has the additional advantage of detecting very low-abundance alleles.
dPCR-based liquid biopsies are still at the translational stage, probably closer to pure R&D at this point than to commercial tests. Developing a liquid biopsy involves first the use of sequencing as a discovery tool to screen the genome for genes associated with the disease of interest. Once the markers are broadly identified, they must be validated not just for “on/off ” but also for relevant levels.
“The body tolerates and can process some positives, thereby avoiding disease,” Banerjee explains. After understanding baseline gene expression levels, researchers must correlate them to the organ and tissue where the disease occurs and determine those cutoffs that indicate true positives from negatives within the context of diagnostics, prognostics, or disease monitoring. This will involve a fair amount of clinical research before one could even consider the regulatory hurdles of having a test and analysis system approved by the U.S. Food and Drug Administration.
Credit: Bio-RadThe competing technologies of Bio-Rad and RainDance illustrate the potential to improve detection sensitivity within a technique. Improvements in primers and reagents are more global in their effect. For example, Transgenomic (Omaha, NE) has developed Multiplexed ICE COLD-PCR (MX-ICP), which enhances the sensitivity of dPCR (and other PCR varieties) by up to 500 times.
MX-ICP is an improvement on COLD-PCR, which enriches variant alleles in the presence of wild-type (“normal”) genes. It works by first targeting a genomic region, then suppressing the wild-type genes and amplifying the mutations. Without the interference from wild-type genes, the rare alleles become more prominent. Company president Paul Kinnon describes the technology as “supercharging” and “platform-independent.”
“When we demonstrated the technology on the Bio-Rad and Rain- Dance dPCR instruments, we had to dilute samples down a million fold to avoid blowing out the detectors.”
Related Article: New Technique Enables High-Sensitivity View of Cellular Functions
MX-ICP was recently tested in a liquid biopsy experiment in which circulating DNA was found to correlate with mutations in the KRAS and BRAF genes, which are common in individuals with metastatic cancer. “If you’re looking for KRAS exon 2, you see the 60 mutations in that region, not just the one, with a limit of detection (LOD) down to 0.01%—defined as the ability to detect one mutation in 10,000 genome equivalents.”
According to Kinnon, greater sensitivity in PCR and sequencing will increase the number of aberrant genes possibly implicated in various diseases. That’s good from a scientific perspective but will require greater validation efforts and disclosures from drug developers. “But if you use that data, and eliminate certain populations from clinical studies, you may not get more blockbusters, but you will get more drugs approved.”
The value of a liquid biopsy becomes evident when compared with a standard tissue biopsy obtained during surgery. Kinnon provided figures estimating the cost of a tissue biopsy at $15,000 to $60,000 when one considers hospital, physician, and ancillary costs. “Tissue biopsies fail about 25 percent of the time because the sample is from the wrong region on the tumor or is degraded during storage.” Plus a tissue biopsy can usually be taken only once. A liquid biopsy is a tiny fraction of the cost of a tissue biopsy and may be conducted as often as desired for prognosis, disease monitoring, or determining treatment effectiveness.