Digital PCR is an end-point PCR method that is used for absolute quantification and for analysis of minority sequences against a background of similar majority sequences, e.g., quantification of somatic mutations. When using this technique, the sample is taken to limiting dilution and the number of positive and negative reactions is used to determine a precise measurement of target concentration.
The digital PCR (dPCR) concept was conceived in 1992 by Sykes et al.1 and then developed into a nanoscale array format by Kalinina et al. in 19972. One of the drivers of continuous improvement of dPCR was the demonstration by Vogelstein and Kinzler3 that rare KRAS mutations could be detected and quantified in material extracted from colorectal cancer patients. Vogelstein diluted the patient samples such that they expected an average of 1 template molecule per 2 wells of a reaction plate. The target sequence around the mutation site was amplified and then two Molecular Beacons (see Quantitative PCR and Digital PCR Detection Methods) were used to detect the amplicons. The first Molecular Beacon was specific to a region that was not expected to contain a mutation and this assay served as a PCR amplification positive control. The second Molecular Beacon had a different dye label and was specific for the wild type (WT) sequence. In this way, reactions showing a positive signal from both labels were expected to be WT and those showing only the positive control were expected to be mutants. Since all the products within a well were expected to be homogeneous as a result of the dilution, an accurate measure of the ratio of mutant to WT could be made. These experiments provide examples of dPCR being carried out in standard 96-well plates, but higher throughput options were suggested by others, including Dressman et al.4 who introduced the concept of using emulsion beads for dPCR (now used in the Bio-Rad QX100™ Droplet Digital™ PCR, ddPCR™ system and RainDance Technologies’ RainDrop™ instrument). In an alternative format, the reactions are run on integrated fluidic circuits (chips). These chips have integrated chambers and valves for partitioning samples and reaction reagents. The first commercial dPCR system using chip technology, the BioMark™, was launched in 2006 by Fluidigm.
When performing conventional PCR, the final concentration of template is proportional to the starting copy number and the number of amplification cycles. One experiment of a given number of reactions is performed on a single sample and the result is an analysis of fragment sizes or, for quantitative real-time PCR (qPCR), the analysis is an estimate of the concentration of the target sequences in the reaction-based on the number of cycles required to reach a quantification cycle (Cq); see Quantitative PCR. When using dPCR, the sample is diluted and separated into a large number of reaction chambers, such that each partition contains either one or no copies of target. The number of reaction chambers or partitions varies between systems, from several thousand when using the QX100 Bio-Rad system to millions when using the RainDrop™ approach. The PCR is then performed in each partition and the amplicon detected using a fluorescent label (see qPCR and dPCR Detection Methods) such that the collected data are a series of positive and negative results. In theory, this would result in a positive signal from a partition that originally contained a single copy of target and a negative (i.e., no signal) from one that did not originally contain template. However, since it is possible that some partitions will contain more than a single copy of template, the dispersal of the sample into the partitions is considered to obey Poisson distribution.
In theory, dPCR can be used to overcome some of the difficulties that are encountered when using conventional PCR. When using dPCR, a sample is partitioned so that individual nucleic acid molecules within the sample are localized into many separate regions and therefore detection of any target is not dependent on the number of amplification cycles. This approach results in a much more sensitive differentiation of fold change than that afforded by qPCR; a well-optimized qPCR may differentiate 1.5-fold changes at best, whereas dPCR has been reported to differentiate 1.2-fold5. The dilution of sample makes this a useful tool for studying minority sequences against a majority of similar but differing sequences. For example, dPCR can be used to detect a low incidence somatic single nucleotide polymorphism (SNP) against a high concentration of WT sequence. This is because, when the total sample is diluted, the rare sequence is also diluted to a single copy, so it will be amplified in the absence of competition from the prominent sequence. Far fewer partitions will contain a positive for the rare SNP than for the WT, so it is possible to make an accurate measurement of the ratio of the two sequences.
Digital PCR has many potential applications, including the detection and quantification of low-level pathogens, rare genetic sequences, copy number variations (CNVs) and relative gene expression in single cells. Clonal amplification enabled by single-step dPCR is a key factor in reducing the time and cost of many of the “next-generation sequencing” methods and hence enabling personal genomics.
Although dPCR is a relatively new technology, many platforms have been developed in an effort to provide better tools for analyses such as; determining absolute quantification without the use of standards, detecting rare genetic mutations using multiplex systems and identifying small fold differences (<1.5) with confidence between diagnostic samples. Across the platforms, the general concept of dPCR remains the same: Input DNA is diluted to generate nanoscale or picoscale reactions containing 1 or 0 copies of template. Individual qPCR assays take place within each droplet containing template, as opposed to a pooled population of target DNA3. In addition to increased sensitivity of detection of rare allelic mutations and CNV measurements, absolute quantification of positive end-point reactions can be made, in part because dPCR measurements do not depend on standard curves or sample calibrations. Other features that can be attributed to increased precision in dPCR quantification include the reproducible and homogeneous droplets generated as well as the increased number of partitioned reactions analyzed6,7. The type of instrument that the end user chooses ultimately depends on the specific application(s) that will be carried out.
One of the first commercially available dPCR instruments was the BioMark™ HD dPCR system (Fluidigm). This platform is based on microfluidic chip technology. The chips can be purchased in a variety of formats, including 12-chamber or 48-chamber arrays. In the 12-chamber array, samples are partitioned into 765 nanoliter reactions yielding 9,180 reactions per chip. The 45-chamber array partitions samples into 770 nanoliter reactions resulting in 36,960 reactions per chip. Samples are loaded into each chamber inlet and nanoliter reactions are partitioned by pressure controlled valves and pumps. Sample partitioning and mixing, as well as thermo cycling reactions are all performed on-chip. Following amplification, fluorescent images are captured using the BioMark system8. On-chip processing allows for less hands-on manipulation, thereby reducing the potential introduction of error while maintaining a more simplified, user-friendly system.
Fluidigm’s dPCR system allows reactions to be multiplexed using four targets per sample. A fluorescent image of the chip is taken before and after each round of thermo cycling. This allows for any pre-thermo cycling background to be subtracted from the final fluorescent image, facilitating accurate counts of positive compartments. Another feature of the system is the ability to quantify template that is partitioned into each chamber using chamber specific, realtime amplification plots. In the event that dPCR is not the only application used in the lab, the BioMark™ HD also functions as a qPCR compatible instrument9,10,11.
The OpenArray® and QuantStudio® 12K Flex dPCR systems (Life Technologies) use microfluidic technology to generate and analyze partitioned samples. The OpenArray system was the first to be developed and can hold up to 3 dPCR plates. The QuantStudio 12K Flex dPCR instrument is able to hold up to 4 dPCR plates. Each 384-well plate has 48 arrays. Within each array, there are 64 ‘through holes’ resulting in 3,072 compartmentalized reactions per plate. Nanoliter reactions are partitioned into the “through holes” using an automated dispensing system and are stabilized there by hydrophobic and hydrophilic interactions between the droplet and the coating on the plate. This platform can be used to multiplex reactions containing two targets per sample. The versatility of this instrument makes it compatible for use in qPCR applications12.
The RainDrop™ instrument (RainDance Technologies) represents another highly sensitive dPCR platform. The increase in sensitivity and quantitative power can be attributed to the smaller, picoliter reactions and the number of droplets partitioned; around 10 million droplets per sample. RainDrop dPCR is based on droplet emulsification microfluidic technology. The chips are designed to hold 8 samples and each sample is partitioned into 10 million reactions, resulting in 80 million partitioned reactions per chip. More input DNA can be included due to the increased number of reaction partitions, making this platform ideal for identifying extremely rare mutations. In accordance with the increased number of reactions, limiting dilution of input DNA becomes less of a concern with this instrument. Furthermore, reports indicate that the RainDrop™ system can be used to quantify 1 in 200,000 mutants and has a lower limit of detection of 1 in 1,000,000. These observations underscore the sensitivity of this instrument13. In addition to enhanced sensitivity, the RainDrop™ platform can multiplex 5 targets per sample simply using red and green labeled fluorescent probes. Varying the amount of fluorescent probe with each target generates a unique color intensity that corresponds to the mutation-specific Dual-Labeled Probe (see Quantitative PCR and Digital PCR Detection Methods) and the intensity is directly related to the concentration of probe used in the assay. The technique of diluting fluorescent probes to generate an optical code may not be limited to a 5× multiplex system, but may be employed to generate a 10× multiplex system, making this one of the most powerful dPCR multiplexing systems available14,15.
Although the RainDrop™ is one of the most cost effective platforms, it cannot be used for qPCR applications. In addition, the set-up is more labor intensive and may, therefore be more prone to the introduction of errors. For example, the droplets are generated in a microfluidic chip, collected and then thermo cycling is performed off-chip. The reaction is then injected into another chip for analysis. When multiplexing samples, droplets containing diluted fluorescent probes are collected from one chip and injected into another chip where they are fused with droplets containing primers, master mix and DNA. The merged droplets are amplified off-chip and then analyzed for the presence or absence of the desired targets15.
The Bio-Rad QX100™ Droplet Digital™ PCR technology is the only platform that does not use microfluidic chips at any stage of the process. Instead, this instrument utilizes oil emulsification technology in a standard 96-well plate format. Eight samples containing master mix, primers, Dual-Labeled Probes and DNA can be loaded into a cartridge at the same time. Each sample is positioned adjacent to a well containing oil and together they undergo droplet emulsification using a vacuum-based droplet generator. Each sample is partitioned into 20,000 reactions. Droplets are transferred from the 8-sample cartridge into a standard 96-well plate. A total of 1,920,000 droplets per plate can be generated for dPCR analysis. Once the droplets have been transferred to the 96-well plate, the samples are amplified using PCR and end-point fluorescent signals are read using a flow cytometric based droplet reader16,9.
The larger number of reactions (20,000 per sample/1.9 million per plate) generated with this platform augments the
precision associated with dPCR in determining absolute quantification and CNV measurements. The increased number of reactions per sample affords the ability to load larger amounts of template DNA when compared to the other systems that have fewer partitioned reactions per sample. This becomes increasingly valuable when detecting rare events of importance. Digital reads from duplicated wells can also be combined to determine CNVs for rare or low copy mutational events. Lower limits of detection have been reported to allow identification of 0.001% of the mutant population in partitioned reactions. Bio-Rad’s QX100™ Droplet Digital™ PCR platform has also been used with maternal plasma DNA to determine absolute quantification measurements of maternal and fetal markers, underscoring the advantages of dPCR technology in measuring CNVs with low quality/low quantity template DNA17,16. Though the QX100™ Droplet Digital™ PCR instrument offers a substantial number of partitioned reactions per plate (1.9 million) and is reasonably priced, it is not compatible with qPCR applications and can be used to multiplex only two targets per sample.
Digital PCR offers many advantages over qPCR. These advantages are made possible by partitioning out individual reactions, thereby enriching low copy and rare allelic amplification, while concomitantly enhancing the precision and quantification power of dPCR as a result of the increased number of microscale reactions. Not only can dPCR be used to measure absolute copy numbers, CNVs and rare allelic mutations, but it can also be used to quantify low quantity/low quality DNA6,16,18,13,19,5,17. For example, when using next-generation sequencing, quantification by dPCR has the potential to eliminate the need and cost of running titration analyses on input DNA. This, in turn, allows the use of smaller amounts of input DNA, minimizing the need for the seemingly biased pre-amplification step commonly used in next-generation sequencing7.
However, there may be some instances where pre-amplification cannot be avoided. When working with DNA from a single cell or when the input DNA is already at a low concentration, whole genome amplification may be required before partitioning the sample into thousands of reactions. It is important to note that pre-amplification of target DNA samples may result in biased amplification of input DNA and this has the potential to skew dPCR results. It may therefore be necessary to examine whether the method used for this step results in amplification bias before assessing absolute copy numbers9,11. In addition, the structure of DNA has been shown to affect copy number measurements in dPCR analysis. This is especially true for circularized plasmid DNA, which should consequently be linearized before use in dPCR applications9.
One of the more obvious drawbacks of dPCR is the initial cost of equipment and ongoing requirement for consumable materials. Relatedly, qPCR instruments are more common in laboratory settings and researchers are more comfortable with handling this platform. Of course, the desired applications of the user will ultimately provide guidance on the type of instrument required to carry out those studies.
Since dPCR samples are partitioned into individual microreactors, the number of partitions determines the range of sensitivity of detection. Quantification relies on counting the number of positive partitions at the end point, as opposed to amplification cycles and therefore does not rely heavily on amplification efficiency. To account for the random distribution of target DNA into partitions, the Poisson statistical model is applied1 and an absolute quantity is calculated. The quantity of the target sequence is typically evaluated in comparison to a reference sequence of known quantity to determine a relative quantification. Applications for the absolute and relative quantification of target DNA include measuring CNVs, biomarker analysis and detection of rare events. In addition, reverse transcription may be combined with dPCR to measure RNA molecules. This technique is beneficial for quantification of low concentrations of virus from complex, transcription in single cells and allele-specific transcription.
CNVs are abnormal copies of DNA due to deletion, insertion, or rearrangement and are associated with susceptibility to disease20,21,22. In cancer, gene copy numbers are often increased and patient responsiveness to drug treatment is correlated to copy number23,24,25,26. Numerous methods have been used to quantify CNVs, including SNP arrays, next-generation sequencing and qPCR. Recently, dPCR has materialized as an attractive tool for quantifying patient biomarkers due to the greater precision in copy number determination. Digital PCR has been used to determine accurately the CNV with a resolution of <1.25. Specific qPCR probes enhanced with the addition of LNA bases (see Quantitative PCR and Digital PCR Detection Methods) may be utilized in dPCR to discriminate and detect the presence of somatic SNP variations (Figure 4.1). Probes are designed such that a mutation-specific probe carries the fluorophore FAM and a second probe, specific to the wild-type sequence (WT), carries the fluorophore HEX. The WT and SNP DNA targets are discriminated in the assay. In some cases, gene copies may be “linked” on the same allele and consequently CNV could be underestimated9. Gene duplication in tandem may be resolved by digestion of template DNA with a specific restriction nuclease surrounding the target sequence16 (Figure 4.2).
Figure 4.1. Detection and determination of the relative copy number of a SNP mutation with Droplet Digital™ PCR. A mutation-specific probe was prepared carrying the fluorescent dye FAM and the probe for unmodified (wild type) sequence carried the fluorophore HEX. The X axis is the HEX signal, generated as a function of the presence of wild type sequence. The Y axis is the FAM signal, generated as a function of the presence of the SNP mutation sequence. Both modified and wild type sequences were detected in the target DNA and the relative abundance of each sequence could be directly determined.
Figure 4.2. Droplet Digital™ PCR assay to determine CNV. Purified, undigested DNA was used as template to quantify gene copy number relative to reference. The same DNA sample was also digested with a restriction endonuclease surrounding the target sequence to elucidate linked copies of target sequence. Poisson error bars are shown.
Digital PCR is used to quantify a variant DNA sequence that is present in a background of abundant WT sequence. For example, somatic mutations that are specific to cancers may be detected when present in a background of normal genotype in clinical samples. Quantitative PCR is limited to detecting mutant sequences present at 1% or greater. Digital PCR provides a tool for sensitive detection of rare copies due to the diluting effect of the partitioning mutant target DNA from the WT. The dynamic range for quantification is determined by the amount of target DNA present and the number of partitions evaluated. Available instruments vary with regard to the recommended dynamic range. The Bio-Rad QX100™ Droplet Digital™ PCR system was used to accurately detect rare mutant DNA from 100,000-fold WT16 and the RainDrop™ instrument was used to detect a mutated copy from 200,000 WT copies13. A typical evaluation of primer/probe sets for use in rare event detection consists of titrating SNP-containing template DNA into WT-template DNA, reducing the SNP-containing template by half at each dilution. An example of such an experiment is shown in Figure 4.3; a single well allowed for detection when the frequency of mutant sequence was as low as approximately 1 in 2,000, while by comparison, the qPCR limit of detection was 1 in 10. The limit of detection for dPCR may be further extended by aggregating data across multiple wells in order to increase the number of partitions without increasing the SNP-containing concentration.
Due to the often limiting amount of DNA sample available for use in next-generation sequencing, the samples are typically amplified by PCR or whole genome amplification. Quantification of DNA molecules post-amplification is critical to the performance of the sequencing assay and could be done with methods such as spectrophotometry. Recently, the capability of dPCR for absolute DNA quantification has been applied to next-generation sequencing library preparation7. A universal template fluorescent probe PCR assay was developed such that a probe-specific sequence is designed at the end of one PCR primer for library amplification27. This universal template probe-based assay may be utilized in conjunction with dPCR to quantify library molecules accurately.
Expression of genes that control cellular activity, including cell differentiation, varies among individual members of cell populations and whole population measurements reflect the average values28. Although it may be preferable to measure transcription in single cells, the amount of RNA present is very small, i.e., <1pg29. The regular workflow for single cell transcriptome analysis using qPCR requires a pre-amplification step to amplify cDNA. Due to the dynamic range and ability to quantify low concentrations of template, dPCR is a suitable method for single cell transcript analysis without the use of pre-amplification.
Figure 4.3. Evaluation of primer/probe set in Droplet Digital™ PCR assay and qPCR for rare event detection. A mutation-specific probe was prepared carrying the fluorescent dye FAM and the probe for unmodified (wild type) sequence carried the fluorophore HEX. SNP-containing template DNA was titrated into wild-type template DNA to evaluate detection of the SNP mutation when present in an abundant wild type background. SNP-containing template was reduced by half at each dilution. A) Primer/probe set was used in qPCR with mutation-titrated DNA template. B) Primer/probe set was used in single well of droplet dPCR with mutation-titrated DNA. The ratio of mutant to wild type copies is shown. Digital PCR allowed for detection when the frequency of the mutant sequence was as low as approximately 1 in 2,000, while the qPCR limit of detection was 1 to 10. The limit of detection for dPCR may be further extended by aggregating data across multiple wells in order to increase the number of partitions.
Previously the “Minimum Information for Publication of Quantitative Real-time PCR Experiments” (MIQE guidelines30) were published to outline experimental design details that are categorized as essential or desirable for publication of qPCR results (see Quantitative PCR). The addition of dPCR as a new tool and the introduction of multiple dPCR instruments, necessitates development of MIQE standards specific for dPCR. The published digital PCR MIQE (dMIQE) proposes essential and desirable elements to consider for validity of dPCR data31. In many aspects, including primer/probe design and optimization, the requirements for dPCR are similar to qPCR.
However, there are properties that are specifically relevant to dPCR. The average copies per partition and partition volume are variable and the values are necessary to apply Poisson statistics accurately and therefore should be reported. Also, the number of partitions from which the results are derived must be documented. It is desirable to include the partition volume variance and standard deviation, as provided by instrument manufacturer. It is essential to include the type and treatment of template DNA used in the experiment. The template DNA is often pre-amplified or digested with restriction enzymes. These methods and corresponding controls must be reported. It is desirable to record optimization experiments, such as temperature gradient and cycle number determinations. The sample volume needed is variable among instruments and therefore is appropriate and desirable to include. It is necessary with all published reports, positive and negative reaction controls and calculated variance and confidence intervals are required. The dMIQE checklist outlines the considerations and designates each as essential (E) or desirable (D).
Since dPCR is still a relatively new technology, there is hope that early adoption of the dMIQE guidelines will prevent the publication of studies that were conducted without appropriate quality and scientific controls.