Should genetic testing go digital?

Keywords::

• digital PCR
• microfluidics
• molecular diagnostics

Increasing need for high-performance PCR analysis

Two scientific milestones in the past 30 years, the invention of PCR and the completion of the human genome, have launched a new era of molecular medicine, one in which the diagnosis, stratification of patients and treatment of disease is increasingly based on measurements of gene expression and sequence. Despite many successes, this paradigm is still nascent, with the full impact likely to be revealed over the next decade or later. Driving this evolution of clinical care is the development of ever more sensitive, precise and cost-effective methods for the analysis of nucleic acids. Enabled by fantastic advances in sequencing capacity, numerous international efforts are in motion to chart the variability and complexity of patients and multiple disease states, most notably cancer. The once fantastical prospect of sequencing every person’s genome is now on the verge of being a practical and even cost-effective proposition. It would seem that the dawn of genomics, heralded over a decade ago, is now finally here. In 10 years from now will we all be sequenced at birth? Perhaps. Will this eliminate the need for targeted genetic testing in the clinic? It is unlikely.

Although genotyping tests may ultimately be replaced by sequencing, the rapidly expanding inventory of disease-specific somatic mutations and gene-expression signatures will drive an increasing need for genetic testing. For the foreseeable future only a fraction of biomarkers are likely to be clinically actionable, putting a premium on the rapid, cost-effective and high-performance measurement and monitoring of small panels of biomarkers. Quantitative real-time PCR (qPCR) has long been the gold standard for such applications and is a cornerstone of clinical screening and diagnosis. Despite advantages of simplicity, low cost, good sensitivity and high dynamic range, qPCR does have important limitations: it does not provide absolute quantification of molecular abundance; even under optimal conditions measurement precision is limited to approximately 20%; and the combination of large reaction volumes, limited specificity, and polymerase errors results in false positives that can severely limit practically achievable sensitivity. These features make qPCR poorly suited to many important applications, including the detection of small allelic imbalances, precise copy number analysis and the detection of rare mutations among a high background of homologous sequence. Is there something better that could immediately replace qPCR in the clinic?

Technical advantages of digital PCR

We believe that newly developed digital PCR (dPCR) technologies now provide a high-performance alternative to qPCR, which should be adopted as the new standard in molecular testing on nucleic acids. In this technique, a sample is first partitioned into multiple reactions at a limiting dilution and then subjected to PCR amplification followed by end point detection to generate a ‘digital’ signal for each reaction, corresponding to the presence or absence of starting template molecules. From this digital pattern, the absolute concentration of starting template molecules may be determined without the need for comparison to a standard.

Although the concept of dPCR was first described by Sykes et al. in 1992 [1], and miniaturized by Kalinina et al. in 1997 [2], it was not until this method was adapted to a 96- and 384-well format by Vogelstein and Kinzler in 1999 that it gained popularity [3]. This is because the power of dPCR depends critically on scale. The precision, specificity and dynamic range of dPCR measurements are tied to the total number of reactions per measurement [4]. In addition, decreasing reaction volumes fundamentally improve single molecule detection by achieving high effective template concentrations, while suppressing nonspecific amplification, a property that is particularly important in the presence of a high, sequence-similar background [5,6]. In addition, given the need for large numbers of reactions, low-volume formats are critical for cost-effective analysis. Finally, dPCR is intrinsically more robust than qPCR, requiring only successful amplification with high signal-to-noise ratio, making it less sensitive to assay design and reaction conditions.

Barriers to adoption

Compared with qPCR, dPCR provides absolute quantification, improved robustness, higher precision, comparable or superior dynamic range and improved sensitivity and specificity. Why then has this method, first reported nearly 20 years ago, not become more widely used? Previously, the technical barriers to adoption were clear: robustness; cost; complicated workflow; and limited access to instrumentation. However, the past 5 years have seen significant advances in the development and commercialization of dPCR platforms that will greatly enhance the use of this high-performance method in both research and clinical testing.

dPCR technologies

As highlighted above, miniaturization is a central theme in dPCR technologies, and available solutions are based either on parallel microfluidic formats or emulsion compartmentalization, followed by serial analysis of droplets.

Emulsion-based dPCR

One class of approaches being advanced for dPCR analysis is based on the amplification of single molecules in oil–water emulsions [7]. This strategy was the basis of the first scalable implementation of dPCR, BEAMing (beads, emulsions, amplification and magnetics) [8], in which the compartmentalization of single molecules with magnetic beads was used to generate bead libraries, having clonally amplified target sequences. These particles were then labeled using sequence-specific fluorescent probes and counted by flow cytometry. The application of BEAMing for molecular diagnostics is currently being commercialized by Inostics [101]. Despite this commercialization effort and several clinical successes, the broader adoption of BEAMing remains hindered by the complexity of the workflow.

Alternatively, the generation of monodisperse emulsions using microfluidic shear flow has been developed for high-throughput dPCR. In this strategy single molecules are isolated in picoliter volume droplets, PCR amplified, and then ‘read’ in serial by flowing them past a fluorescence detector. Two recent reports [9,10] have used this approach to achieve high-performance dPCR and these systems are being independently commercialized by RainDance Technologies [102] and BioRad [103]. The RainDance ThunderStorm™ platform, expected to be launched in early 2012, is based on a microfluidic format for pL droplet formation and reading. The recently released BioRad-QX100™ system, developed by QuantaLife/BioRAD, uses a 96-well plate format with 20,000 1-nl droplets analyzed per sample. This implementation may be well-suited to applications where low cost and variable sample sizes are important.

Chip-based dPCR

The most proven dPCR platform is the Fluidigm Digital Array™ [104], which uses integrated microvalve technology for partitioning planar arrays of nanoliter-volume reactions on a microfluidic chip [11]. Major advantages of this system include robustness, extremely simple workflow and the ease by which multiple samples can be analyzed in parallel, making it ideally suited to higher throughput applications and central testing. The current version of this system (48.770 format) allows for the parallel analysis of 48 samples with 770 reactions per sample, each assayed in 850-pl volumes. Our group has recently used surface-tension partitioning, rather than mechanical valves, to extend this density, performing chip-based dPCR at the scale of millions of reactions per run, achieving reaction densities of 440,000 cm^-2 and reaction volumes as small as 3 pl (this technology is being commercialized by Fluidigm) [4]. This ‘megapixel’ format conserves the simplicity and workflow advantages of chip-based dPCR while achieving performance comparable or better to the best reported for droplet formats; a dynamic range up to 10⁷, single-nucleotide-variant detection below one copy per 100,000 wild-type sequences and the discrimination of a 1% difference in chromosome copy number. In addition to enabling very demanding diagnostic applications, including tests based on precise copy number analysis or the detection of subtle allelic imbalances, large dPCR arrays may be split over many samples to further increase throughput and reduce cost.

Other emerging chip-based solutions for dPCR include the OpenArray™ (Life Technologies) [105] and the SlipChip (SlipChip LLC). The OpenArray™ platform, initially developed by BioTrove for qPCR applications, consists of 3072 33-nl reactors, created by microfabricated ‘through-holes’ that use hydrophobic coatings to immobilize droplets. In this approach, automated dispensing is used to load the reactor array with a total throughput of approximately 36,000 reactions, and up to 48 samples per run. Most recently, Kreutz et al. described a SlipChip dPCR device that uses the translation of two microfabricated substrates for the compartmentalization of reactions; samples are loaded into oil-filled channels created by matching features on two glass plates and the chip is then ‘slipped’ to isolate defined volumes [12]. Despite having a modest number of total reactions per device, this system uses a multivolume approach to preserve good dynamic range and sensitivity.

Applications in molecular diagnostics

Regardless of the technical implementation, dPCR is likely to become a cornerstone of molecular diagnostics. In addition to the unmatched technical capabilities described above, absolute digital quantification greatly simplifies the interpretation of raw results and facilitates more standardized clinical interpretations and comparisons across patient cohorts [13,14]. Although dPCR has yet to be adopted in routine clinical testing, several studies have already proven the power of this approach in demanding diagnostic applications, including the detection of rare mutations and small allelic imbalances.

The first major application of dPCR has been in the detection of low-level mutations in cancer. In one example, the BEAMing method was used for the quantitation of patient- and tumor-specific mutations in circulating tumor DNA in subjects with colorectal cancer [15]. Digital measurements of circulating tumor DNA levels outperformed carcinoembryonic antigen analysis for predicting disease recurrence following surgical resection. In a separate study, BEAMing was extended to methylation analysis and applied to early detection and disease monitoring of colorectal cancer [16], with a reported increase in sensitivity of approximately four-times higher than that obtained with carcinoembryonic antigen alone.

The most widely used method for scalable dPCR is the Fluidigm platform. Wang et al. showed that the increased sensitivity of this platform allowed for the detection and quantification of EGFR mutations present in early-stage resectable non-small-cell lung cancer tumors and prophesized that the early identification of such mutations may allow earlier therapeutic treatment with curative intent [17]. A separate study compared the performance of this platform with nucleotide sequencing and allele-specific oligonucleotide PCR in the detection of an ABL tyrosine kinase domain mutation [18]. Although the three compared methods agreed in the majority of cases, dPCR was shown to enable detection earlier than sequencing approaches.

In addition to the detection of rare molecular events, the precision of dPCR is uniquely suited to the analysis of copy number variations and small allelic imbalances. One study used the Fluidigm platform to assess ERBB2 copy number variations, a marker shown to be associated with herceptin response, and showed that dPCR measurements corresponded to qPCR measurements, but had significantly less technical variability [6]. In a separate application, dPCR was used as a rapid alternative to replace cytogenetic analysis for fetal karyotyping from amniotic fluid and chorionic villus samples [19]. dPCR has also been used to accurately quantify the amount of fetal DNA circulating in maternal plasma [13], with implications for the analysis of subtle allelic imbalances as a noninvasive diagnostics of fetal aneuploidy [20], an approach that has been validated using next-generation sequencing [21].

Regardless of the technological implementation used, all of these studies highlight the enormous potential of dPCR in diagnostics, not only for improving the performance of existing tests, but also for enabling new applications that are otherwise inaccessible. Although regulatory approval may present the most significant remaining barrier to clinical use, this process has already begun, and the availability of well-established dPCR platforms has paved the way to widespread adoption of this technology. When coupled with the rapid advancement of new biomarkers, dPCR will find numerous high-impact applications, including the early detection of cancer, the monitoring of chemotherapy response and residual disease, the identification and quantification of pathogens, prenatal diagnostics and immunology. Continued technological advancement will expand the pervasiveness of dPCR, with new applications ranging from rapid point-of-care testing to scalable single-cell analysis. Given the increased pace of progress over the past few years, this may not take long. In any event, it would seem that the future of genetic testing will soon be written in 1s and 0s.

Financial & competing interests disclosure

CL Hansen is a member of the scientific advisory board of Fluidigm corporation and has a financial interest in this company. Fluidigm has provided funding for research in the Hansen group, including the development of high-density digital PCR. CL Hansen has financial interest in a pending patent related to the subject matter of this review, US patent application #20110053151, for a microfluidic device and method of using the same (2011). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.