NAME: Hanlee Ji
POSITION: Attending Oncologist, Veteran’s Administration Medical Center, Palo Alto; Senior Associate Director, Stanford Genome Technology Center, Stanford University
BACKGROUND: Assistant Professor, Medical Oncology, Department of Medicine, Stanford University School of Medicine
Digital PCR, especially droplet-based platforms such as those offered by companies like Bio-Rad and RainDance Technologies, is emerging as a useful tool to accurately detect and quantify rare genetic events and DNA copy number variations.
The research group of Hanlee Ji at Stanford University has been studying copy number variation in cancer tissue samples, specifically gastrointestinal adenocarcinomas. Ji and colleagues recently adopted a Bio-Rad QX100 Droplet Digital PCR system to help them more accurately detect and quantify amplifications of the fibroblast growth factor receptor 2, or FGFR2, gene, which has been shown to undergo amplification in gastric and breast cancers, among others.
In a paper published in September in the open-access journal Translational Medicine, Ji and colleagues described their use of droplet digital PCR to detect FGFR2 gene amplifications in formalin-fixed, paraffin-embedded tumor samples, and demonstrated that the technique was more accurate than quantitative real-time PCR assays, which can be confounded by FFPE tissue samples.
Ji took a few moments this week to discuss with PCR Insider his group's work and the burgeoning utility of droplet digital PCR in copy number variation studies in cancer research. Following is an edited transcript of the discussion.
Your lab is involved in many different genomics projects. Provide me with some background on your work in genomic amplifications.
There are three arms to my research group. First, we're interested in developing genome sequencing technology. Not only are we early adopters, but we've developed a lot of our own technology in house, [and] we routinely use these tools for cancer genetic translational studies. The majority of this has been built around next-generation DNA sequencing, but increasingly we are using droplet digital PCR, among other systems, as a way to improve sequencing or as an independent assay for cancer genetic studies.
We also have a bioinformatics part, because we do so much next-gen [sequencing], so a portion of the group is really focused on developing new statistical algorithms, pipelines, and what not.
And the third component that I think distinguishes us from many groups is, given that I'm a practicing medical oncologist, that we have translational cancer genetics projects, where we apply a lot of these and our own technologies to try and analyze these patient populations to determine if there are [discoveries] of clinical utility that would have some long-term benefit for these patients.
Prior to implementing droplet digital PCR, what were the most commonly used analytical methods in your lab for quantifying genomic amplifications?
We use primary genome sequencing, whether it's whole-genome; low-coverage whole-genome; high-coverage; exome-based; or targeted resequencing-based among the next-gen approaches. We also use arrays, and we've had quite a bit of experience with traditional quantitative PCR.
Each of these methods has certain drawbacks and advantages, but the paper suggests that one major issue was not being able to accurately measure these amplifications from FFPE tissues. Why is the accuracy of these amplifications so important?
There are two aspects that are important. First, trying to detect these kinds of events from archival cancer samples is not trivial. We wanted to have a tool that would be relatively robust, particularly for DNA extracted from archival material. Clinical samples are most often embedded in paraffin and have been treated with formalin, so that tends to create a number of problems with the DNA itself.
What made digital PCR so appealing is that it has a robustness in terms of performance, ease of use, and, as that paper really focused on, nice applicability to DNA samples taken from FFPE tissues, which really opens the door to doing larger translational studies in clinically annotated patient populations.
Can you explain in a little more detail the main finding of the paper, how droplet digital PCR compared to qPCR for detecting genomic amplifications.
This is fundamentally an issue dealing with interpretation of signal from quantitative PCR using traditional methods as opposed to using digital or counting methods … We found that [ddPCR] was more accurate, particularly based on our control samples, and I think from my perspective as someone who runs a lab that the actual practical part of setting up the assays, the general robustness of performance, and the ease of use was such that digital PCR was overwhelmingly much more robust and simpler to adopt. It just showed general improvements in all aspects of performance.
In all my experience with qPCR – and we have another study we've published on acute lymphoblastic leukemia where we had to do some degree of qPCR – is that there is too much voodoo involved in trying to get that to work routinely. We found digital PCR to be a beautiful complement to our next-generation sequencing, either as a validation tool or something to use once you've identified an interesting phenomenon and you're checking in larger populations about the prevalence of those events.
I have to ask you to expand on what you mean by 'voodoo' with qPCR. Is it a matter of it being open to interpretation because it's not an absolute measurement?
Absolutely. It almost always has to do with the type of controls you're using and how you ultimately quantitate things. Whether you're running a series of standards alongside it, or whether you're using internal controls – I've just found the interpretation of that to be fairly nuanced at times, and have never been very happy about the methods that are required. There is a fair amount of tweaking that needs to happen in order to get it to work, and you have to make very careful decisions about the type of controls that you need to run in order to make it interpretable. We found out that a lot of that we don't need to worry about when we do our droplet digital PCR. We don't have to include so many standard controls or standard curves, in particular. That's reduced the work and the complexity, and made it much more reproducible.
Do you think ddPCR works better than qPCR in other tissue types besides FFPE tissue? For instance, fresh or frozen tissue?
Yes, it works even better. The higher quality the DNA, the better quality of data. Another thing that I've found easier to do is that the readouts that we see from digital PCR are fairly specific, and we've generally found it easier to troubleshoot because we are able to look at the individual points that are being counted and get a sense of those distributions visually and graphically, and also in terms of the way it's being analyzed. It's just a much more transparent system, to me, in terms of how to assess that information as it comes off the digital PCR system.
What are the main limitations or drawbacks to using droplet digital PCR right now?
I think that the improvements include increasing the throughput, so being able to do it on plate-based systems. Our current instruments don't currently have the capacity to do 96 wells. Expansion of the type of enzymology that can be used within droplets – we're pretty limited in what types of enzymes can be used. At least the current iteration uses Taq polymerase, and we'd certainly like to use different types of polymerase. I know that's in the works, but certainly a wider range of different enzymes that can be applied to all different types of assays would be helpful. An inherent limitation is that the amount of multiplexing that one can do within a given assay that's defined by a different well is limited — so if there are ways to begin to use multi-color, to try and expand the multiplexing, that might become an advantage later. It would certainly make the instruments more expensive, but as more of these get released, maybe there will be an opportunity to incorporate sensor modules that can discriminate different colors.
What's next for your lab with this work? Do you need to test digital PCR against other methods for certain assays still? Also, given this information you found on genomic amplifications in FGFR2, what are you doing with that data?
The two primary tools that are used in my lab are NGS for cancer genetic studies, and increasingly using digital PCR to both validate those findings and conduct separate analyses of particular events that we're interested in.
One expansion of the study that is underway and should be completed within the next couple of months is that we're going through a fairly large set of clinically annotated gastrointestinal tumors and identifying therapeutic targets denoted by gene amplifications systematically. And we're doing that in fairly rapid order. I think the longest lag time is coming from optimizing primers. But the study, in terms of looking at large numbers of samples, can go quite quickly. That data is obviously fitting into a large translational study of this particular type of gastrointestinal cancer [where] we're trying to identify both prognostic events that may have implications regarding survival; as well as to try and do some predictive analysis, where we try to identify personalized drug targets by using digital PCR and NGS.
Do you think that digital PCR has the potential to be used as an analytical method in a clinical sense, for actually detecting these amplifications in real time from patient samples in order to personalize treatment?
I think the adoption in terms of doing personalized cancer genetic analysis … off of specific loci that have clinical ramifications – I anticipate that digital PCR will be widely adopted in the next couple of years. And eventually, I think there is the opportunity to completely replace quantitative PCR as it is done now.