DNA sequencing has become a go-to method for detecting genetic mutations in certain cancer tissue to determine whether patients will respond to a given treatment. However, it requires an investment of time and money that is often at odds with the resources of a typical pathology laboratory.
Researchers have recently demonstrated that combining PCR with high-resolution melting analysis is an inexpensive and reliable alternative to sequencing for screening for such mutations. However, this method is limited by the quality of the DNA template — which is often poor in the formalin-fixed, paraffin-embedded tissue samples typically processed in pathology labs.
To address these issues, researchers led by Mohammad Ilyas of the Queen's Medical Centre at Nottingham University in the UK recently developed and validated a technique that combines quick-multiplex-consensus PCR with HRM to detect mutations in FFPE tissue.
Their work, published in the February edition of the Journal of Clinical Pathology, demonstrated that QMC-PCR with HRM was able to detect a minimum of 2.5 percent of mutant alleles — compared with 20 percent for Sanger sequencing — in DNA from FFPE colorectal cancer tumors.
The researchers claim that the technique may be an inexpensive, fast, and easy way to screen for mutation hotspots in multiple genes from FFPE biopsies in order to profile tumors and help guide diagnosis and therapy selection.
PCR Insider discussed the development of the assay this week with Ilyas. Following is an edited version of the interview:
What was your motivation for conducting this research?
I'm a histopathologist by training, which means my stock in trade is looking down the microscope at tissue. Somebody has a biopsy taken, it comes to the histopathologist, we look at it and make a diagnosis, and then the clinicians act on our diagnosis.
But I also have a research interest in the genetics of colorectal cancer. The exciting thing about [cancer] management in this day and age is that the genetics are becoming more and more important. We can now tailor specific therapies based on the genetics of a tumor, because we know that the presence of a mutation will predict whether somebody responds to a specific therapy or not.
A major advance in colorectal cancer has been the identification of the fact that mutations in the KRAS gene will predict whether someone responds to anti-EGFR therapies. EGFR is a cell surface receptor, and there are two main therapies: cetuximab [Erbitux, manufactured by Bristol-Myers Squibb and ImClone] and panitumumab [Vectibix, manufactured by Amgen], both of which are monoclonal antibodies. In a certain number of tumors, they work very effectively, and in a certain number of tumors they don't. The difference in the tumors that don't respond is the presence of the KRAS mutations.
The drive now is to identify tumors that have KRAS mutations to decide whether they are appropriate for this therapy or not. There are some challenges in that … there are three different hotspots for KRAS mutation: codon 12 and 13, codon 61, and codon 146. Not only that, but there are other genes in the KRAS signaling pathway that may also be mutated — BRAF, for example. And because BRAF lies in the same signaling pathway, it is possible that someone could be non-mutant for KRAS, and you think they're suitable for [anti-EGFR] therapy, but if they've got a BRAF mutation, they won't respond to that therapy.
So predictive testing is complicated by the fact that you may need to test for more than one gene, because it's actually a signaling pathway. That means multiple hot spots may need to be tested. The only way you can be absolutely sure whether something has mutated or not is by sequencing. But sequencing is quite expensive. If you can screen beforehand, then you reduce the amount of sequencing you have to do.
For example, patient A has a colorectal tumor, and we want to know if he has a mutation in the KRAS pathway or not. We have to test at least five hotspots in that pathway: three in KRAS and two in BRAF. Rather than sequencing five different PCR products, it is easier to screen the ones that are mutants, and you can verify the presence of mutations by sequencing just the one that has come out positive in your screen.
High-resolution melting is just a beautiful method for screening for mutations. It was originally developed as a method for looking for SNPs, but it is also good for looking at other mutations, like missense mutations. It will also pick up on small deletions, as well, because it depends on the formation of heteroduplexes. This is why we started looking at HRM. We were looking for a way of practically screening tumors, because we needed to screen more than one gene per tumor.
How are these mutations tested for currently? Aren't there tests on the market? The KRAS mutation kit offered by Qiagen subsidiary DxS comes to mind.
The DxS test is based on Scorpions probes. It's a real-time assay and it's exquisitely sensitive. The problem is that it's sequence-specific. That means that for one of the hot spots in KRAS, I think there are eight different probes, and each probe is sequence-specific; whereas with the HRM test, you can screen for any mutation, because any mutation will cause heteroduplex formation. That means that we can screen one of the hotspots in KRAS — let's say codon 12-13 — using just one pair of PCR primers; whereas the DxS test needs eight PCRs for that. And that's just one hot spot. Then if you go to KRAS codon 61, DxS doesn't have a test for that. When they do, it will be another series of primers and probes, because there are several sequence variants that can be mutated there. And when you get to codon 146, there are several [variants] there. With HRM, there is only one primer pair for each hot spot. The DxS test, in some ways, is just too specific, which means you have to do multiple tests to screen for all possible variations at that site. HRM just provides a lot more flexibility.
Previous research had shown that HRM could be used to detect these mutations, but your paper demonstrated how it could be used with formalin-fixed, paraffin-embedded tissue, correct?
That's right. What happens currently with tumors is that a patient has the tumor taken out, and usually it's put in formalin, which fixes the tissue. That allows us to preserve it so we can look at it under the microscope. The problem with formalin is that it causes cross-linking and fragmentation of the DNA.
Whenever you extract DNA from fixed tissue, the quality of the DNA is much poorer than fresh tissue. But, the vast majority of specimens — certainly here in the UK — are fixed in formalin. We've got a PCR-based test, but we've got to get it to work using a rather poor template, because the DNA is so degraded. We've found that up to about 300 to 400 base pairs is just about OK for PCR, but beyond that, in FFPE tissue, it becomes highly unreliable.
With HRM, what has been reported in literature and what we were finding is that when you start off with a poor template, you get a high frequency of false positives. When dealing with very high-quality DNA, from cell lines, for example, it's an absolutely perfect procedure. When you're dealing with low-quality templates from FFPE tissue, you get a false positive rate of about 15 to 20 percent. We would find aberrant melting, but no mutation. We were faced with the problem of too many false positives. So we had a test to screen the hotspots to reduce the amount of sequencing we have to do; but on the other hand, it was producing so many false positives, we were losing the benefit of screening.
This is why we went down the nested route. Nesting is a two-step reaction that just gives you more specificity. We found that we could split the procedure into two reactions: a first reaction where we had all the hot spots together, and that's what we call the pre-diagnostic multiplex reaction. And then we could take that PCR product, dilute it by 100, and then run specific PCRs on that product. That's what we call the single specific diagnostic reaction.
I feel that we've now got a very flexible technique where we can do one PCR reaction to amplify all the hot spots together. We tried it with up to 10 primer pairs, so we can amplify it at 10 hot spots. And then we can do a specific diagnostic reaction [that] is suitable for HRM analysis. We've gotten over the problem of the false positives because of this nested reaction. We're getting a much purer PCR product to actually do the HRM analysis. But because we do the initial multiplex test, it allows us to do a lot of things in one test. It economizes the use of DNA templates, because you don't have to use as much DNA, and when you're dealing with smaller biopsies, the amount of DNA may be a limiting factor.
The other benefit is that we are just using a single PCR program for both sets. It's just a two-cycle thermal cycle, so it's quite quick. We feel we've killed a number of problems with one test.
Have you confirmed why the test works with FFPE tissue? You hypothesize in the paper that the reason that the nested technique works is because contaminants are diluted out and only specific PCR product is used as template.
We've sequenced all the diagnostic PCR reactions. We think that's why it's working, because it's using enriched template and giving a single, specific product.
The thing is, if we were doing straightforward sequencing, I would probably only use a single set; I wouldn't use the nested procedure, because in sequencing, if your primers are good, it's fairly reliable. Whereas with HRM, if your primers are good, but you get a little bit of background, it just screws your HRM up. This is why it needs the two-step reaction: to get rid of that small amount of background. Any contaminant will just cause aberrant melting. This is what allows us to have good HRM analysis, because the first round of PCR, the pre-diagnostic multiplex reaction, just enriches the hot spots; and then the diagnostic reaction amplifies those.
How would this technique be employed in a histopathology lab? Does it require some sort of regulatory approval?
This complements the diagnostic work we do. We actually make the diagnosis on the microscope, and the genetic testing allows us to define what kind of cancer it is. Then we can tailor the treatment based on the genetic profile.
With regards to regulatory approval, the law is a little unclear in the UK. In Europe you have CE certified tests, but within the [UK's National Health Services], there is scope to make your own test, and to use homebrewed kits. For example, with the KRAS testing, we can use our own test if we feel it is sufficiently reliable. There is no legal obligation to use CE marked tests.
Are any of the methods you developed patentable?
We looked into this. Because for the QMC-PCR we are just using well-established techniques, like HRM and nested PCR, there is nothing really patentable. We've taken what was there and modified it for our purposes.
We are trying to develop other techniques based upon HRM, which probably won't be patentable, because again, they're well-developed techniques. At the end of the day, our lab is a research-first lab where we're more interested in the biology of disease, and less interested in method development. It's just that we had to develop this method because it wasn't quite good enough for FFPE tissue. It was all driven by necessity.
The thing that excites us the most about this is we've got a method where we can screen a large number of different hot spots on a DNA template that is quite poor. We get good quality data from rather poor starting material. And we feel that it is quick and easy — about an hour and a half; and not particularly expensive — it's just two PCR reactions, and uses existing laboratory equipment. Once it's been validated by a number of different studies, I think it will easily be applied within the healthcare sector.