There has been a lot of excitement around liquid biopsy as a way to diagnose and monitor cancer. Now it appears that all the hype has some genuine clout. Last month the Dana Farber Cancer Institute demonstrated that liquid biopsy, in the form of droplet digital PCR from Bio-Rad, can rapidly detect gene mutations linked to non-small cell lung tumours, giving clinicians a better shot at prescribing the right course of therapy. These findings are so promising in fact that DFCI is planning to role out the test to patients outside of their clinical trial.

Bio-Rad are rapidly making a name for themselves on the liquid biopsy scene, supporting clinical studies like that at DFCI and making moves into the European clinical market. We sat down with George Karlin-Neumann, scientific director of affairs at Bio-Rad’s Digital Biology Center, to get to grips with what droplet digital PCR is, why it’s making waves in the world of cancer diagnostics, and where the future developments are.

FLG: Firstly, could you give me some background to the technology that was used in the study?

GKN: This is a droplet form of digital PCR, as opposed to other types where fixed physical nano-chambers are used. In this particular droplet digital PCR technology from BioRad there are about 20,000 partitions. The sample is made up in a PCR mix that’s suitable for stable droplet formation, and then is put into a disposable cartridge that will take eight samples at a time and emulsifies them all in a couple of minutes. This is done in a droplet generator instrument that preps the samples for transfer to a PCR plate, which is then thermocycled in a standard thermocycler. Subsequently the plate is read in a droplet reader, essentially a droplet flow cytometer, which enables you to autosample each of the 96 wells and to read the droplets single-file from each sample, and measure in two fluorescent channels whether a droplet has zero, one or both of the two fluorescent signals. So essentially any droplet that has a target of interest in it will develop a strong fluorescent signal and will be recorded as a positive for that particular channel.

What you come out with is a population of droplets for each sample, which consist of a fraction that are negative droplets and a fraction that are positive droplets, and to a first approximation for a very low DNA input, the number of positive droplets you have roughly corresponds to the actual number of target molecules of interest in the volume of sample that you’re analysing. So from that you can calculate via Poisson statistics what the concentration in your real biological sample was.

Often less than 10% of the droplets are actually occupied by a target of interest. However, one of the things that’s worth stressing in the BioRad system is that because of the extremely high uniformity of the droplets, and each droplet’s grabbing an equivalent portion of your total sample, you can actually load these up 50-100 times more heavily than in that limiting dilution regime, so that you can load in enough DNA to allow you to interrogate 100,000-150,000 genome equivalents per sample reaction without saturating all the droplets. To be able to know what your concentration is you need to have at least some fraction of negative droplets, and without that you’re without your moorings, you have no idea how far you are beyond saturation, so as long as you have some negative droplets you can use the Poisson statistics to calculate back what original concentration of sample would have given you those fractions of negative and positive droplets, assuming equal distribution of your sample.

FLG: What are the key differences and advantages does ddPCR have over technologies like NGS in a clinical setting?

GKN: The way I like to think about it is in terms of asking about clinical care of a patient. Within oncology you have to ask the question about the complexity of the disease inherently, whether it’s melanoma, lung cancer or blood cancers. How many different potential drivers there are of that disease, and how much information do you need to acquire about that particular patient at the particular point in the trajectory of their disease? These are the key questions you have to ask yourself.

In some cases there may be many drivers, such as in breast cancer, and you may not know in a given patient what’s driving that disease. So you need to genotype the cancer, and you may need to be open to the possibility that there are a number of different genes that could be driving it and are therefore most suitable for an intervention, for say a targeted therapy. In those cases, something that has a broader profiling ability such as next generation sequencing is probably the best technology at this time that we can apply to that. However once you’ve done your initial genotyping of the tumour, through either tissue or blood, if you want to actually ask whether or not the tumour is responding to treatment, shortly after the start of therapy, or after surgery, or monitoring for months or years later, and you know what you’re looking for, then it is much more expedient, much quicker and at this point more sensitive to use a monitoring method like digital PCR. You can develop assays that interrogate specific sequences, such as the driver oncogene sequence mutations.

So in that case monitoring is, I think, a real strength of using a technology that’s well-focussed and highly sensitive, quick and far less costly. In the course of disease it appears that there is a very strong correlation between the level of tumour DNA sequence in the blood cell-free DNA, and the size of the tumour. So liquid biopsy implemented with ddPCR offers a very good surrogate for that.

However you may find that the tumour escapes from the selective pressure of the treatment and acquires a new mutation that continues to drive tumour growth even in the presence of a therapy that was previously working. At that point it is prudent to say what is now driving the disease, and do we have a different intervention that can help to address that. At that time it may be fruitful to do a broader profiling, pulling in NGS again. So I really see this as a back and forth between the technologies with respect to the course of disease, what you know about the patient’s disease and what information you need to know.

Ominous cancer cellsFLG: The Dana-Farber Institute is now planning to offer the test to eligible patients outside of the trial. What will be the main challenges in extending the ‘real-world’ reach of this test to clinicians and patients?

GKN: The real world challenges are technical and logistical, which were well-addressed in the paper. Anything from drawing a sample, where you draw the sample, how stable is the genetic information in that sample, transporting the blood tube, how quick each process is etc. So there are these kinds of logistical issues, which were really addressed head-on in the Dana Farber, so they really could say if we bring this into a clinical setting, what are the issues we face? They drew the contrast between evaluating the sample with ddPCR versus Next Gen Sequencing.

I think there are regulatory issues that are important to whether a method can be adopted in a particular geography, whether it be the US, or Europe, or Asia, and whether the system needs to have been approved by a regulatory body in any of those locations. BioRad is currently moving towards getting the QX200 system cleared through the FDA, and also in other jurisdictions to follow.

Then I think there is a third aspect that’s key, which is does it do any good? Does the information that you’re generating actually improve the care of patients? People are very much expecting the answer to be yes, but I think when you really come to the point of asking health providers, insurers to actually pay for these tests, they need to be convinced that they actually help with patient treatment.

FLG: How could droplet digital PCR be clinically applied beyond cancer diagnostics?

GKN: The real frontier to my knowledge is transplant monitoring. The donated organ, whether it’s a kidney or a liver or a heart, can essentially be seen as a foreign genome being transplanted into a host, and by knowing distinguishing genetic markers in the donor’s tissue the abundance of DNA from the donated organ in the blood can be used as a measure of how rapidly that organ is turning over. There have been a number of digital PCR and Next Gen Sequencing studies done that show that during the first week after transplantation there is a huge contribution of the donated organ DNA in the blood stream, up to 90% of the cell-free DNA in the first two days. This rapidly drops down within a week to maybe several per cent or less depending on the size of the organ that’s been transplanted. One can conceivably use that monitoring to find out whether the organ is in jeopardy, through a direct measure of organ turn over, similar to measuring changes in tumour size. That is being seen as a tool to allow more direct assessment of whether treatment with immuno-suppression is being overdone or underdone in the patient.

Another area that potentially shows promise is pre-disease screening of Type I diabetes. There are several published studies showing that there are markers for a gene that is methylated specifically in the islet cells in the pancreas, and that same gene sequence appears not to be methylated in other tissues in the body. So in the healthy patient there will be a very small fraction of methylated insulin gene DNA in the blood, but it appears that before clinical signs of diabetes are evident, if the islet cells are starting to turn over rapidly due to disease development that signal can be detected in the blood, and could potentially alert us earlier that an intervention needs to be made before the complete loss of the islet cells. It will be interesting to see how robust those findings are, and whether this can be introduced into clinical practice in the future.

Another application is non-invasive prenatal diagnostics. One example is assessing whether the developing foetus has a chromosomal aneuploidy, which is often done using Next Gen Sequencing but can also be done by digital PCR by detecting markers on a potentially aneuploid chromosome versus a reference chromosome.

The last thing is something that I have seen a single report on, which is to detect sepsis at the very early stage. To my mind those are some of the opportunities in the future, but transplant monitoring is to my mind the most extensively studied so far outside of oncology.

Thank you for your time!