Role of ctDNA in Germline Testing

Mark Robson, MD:I want to touch a little bit on this new technology and get your vibe on how you are handling cell-free DNA (cfDNA) profiling. Of course, there are a number of different companies that are doing this now, so-called liquid biopsy or whatever you want to call it. And it's not infrequent that you find low–variant frequency alterations in cancer susceptibility genes. You get your cfDNA test, and there's a 0.2% BRCA2 variant or something like that.How do you go about that? Dr Ross, let me pick on you first.

Ashley Ross, MD, PhD: Well, there are a couple of considerations. This comes up in prostate cancer, as there's sometimes bone-only progression. There is nothing easy to biopsy. And you're now trying to think whether you can assign this person a therapy. And with BRCA2, as you mentioned, you think you can now qualify them for a PARP [poly adenosine diphosphate-ribose polymerase] inhibitor, but is that their next line or not? But could this be spurious? Could this be hematopoiesis of uncertain significance and just something that happened in a CHIP [clonal hematopoiesis of indeterminate potential] phenomenon? Could this be real? Should it guide treatment? When you don't have options, you use it as leverage to get the drug you want through clearance for insurance. When you have other options, you kind of think, “Well, I'm not really sure. I've had no guidance on how to interpret the allelic frequencies of cfDNA.” I would say I'm often bamboozled in using it the way I want to leverage the next line of therapy. So that would be my 2 cents. And now, to the more sophisticated answers from Dr Henson or Dr Clark.

Mark Robson, MD: It is a complicated issue, because whether or not you're finding subclonal alterations, whether or not you're just finding CHIP … our assay happens to be also a plasma/buffy coat assay, so we subtract CHIP from ours. But that's not the case for most of the available assays, and it is a challenge. Dr Henson, we were talking about F1. What do you think about the issue of mosaicism that you're picking up on ctDNA [circulating tumor DNA] or cfDNA?

John Henson, MD: In most hereditary cancer situations, mosaicism is not as big a phenomenon. What's even worse than an F1 [first filial generation] is an F2 [second filial generation], where mosaicism is a huge issue. When you see a variant allele fraction in plasma that is on the order of a few percent to 15%, that's very common with what you see with ctDNA. CHIP is supposed to be less than 20% or 30%; mosaicism is under 30%. Mosaicism in most solid-tumor situations is actually rare, and I would encourage people not to spend too much time worrying about mosaicism unless there's some reason. With CHIP, on the other hand, the genes involved are fairly well known. If you have a tumor that has an APC variant at 30%, that's not a common CHIP gene, so I'm not thinking CHIP in that situation. And then you really have to think about is whether this is something that's being shed out of the tumor. If the level gets up to 50%, obviously, it's different. But there again, and I think Dana mentioned this, the variant allele fraction (VAF) is very tricky, because it could be that you've lost the other allele, and now you're only amplifying the 1 allele that's mutant. But if it does go up to the 50% range, then you need to think about the possibility that the patient has a germline change. Obviously, the canonical percentage would be 50%, because 1 allele of all the lymphocytes contributing to the circulating DNA is going to be abnormal. So low frequency, it's probably going to be a tumor. There are all kinds of problems with these assays, like low metastatic burden. Certain tumors don't really shed that much tumor DNA.It's helpful in really advanced cancers. And in the early stages, I think we have more work to do. Once we get up to 1000x coverage, we may get a lot more information. But anyway, it's a complex subject.

Ashley Ross, MD, PhD: And just so it's clear, and not to poo-poo cfDNA, I found it very useful in prostate cancer. We're often targeting the antigen receptor (AR), and we will sometimes see cf DNA emergence, resistance, or alterations in AR that are actionable, so that's very helpful. I've also been very impressed with how much they've been able to push the envelope on cfDNA, such that they can even start to report things like if they're not MSI-high [microsatellite instability-high], you won't really know. Sometimes they're doing a 100-gene panel, and they can start to look at some of the MSI or even at beta testing or tumor mutational burden. There are certain things to learn about resistant clones that might be out there and about some of this immunogenicity. And that really speaks toward the idea of doing serial testing throughout the patient's treatment lines, so I think there is a role. So just for the audience, there are caveats, but I'm glad that it exists, and it has a role in patient care.

Mark Robson, MD: Dana, have you seen cases referred to you on ctDNA?

Dana Farengo Clark, MS, LCGC: Yes, we sure have.Penn Medicine uses Guardant [Guardant Health, Inc] a lot, and so what we've done is partner with the Guardant people in our region, because the ability to get any germline information from a commercial liquid biopsy lab is pretty difficult. We've asked them if they could provide us with any additional information. They will send ones that they think they can't put on a report and send us a little heads-up email saying, “We think you might want to bring this person back in.” We're trying to backdoor it a little bit. But I do think the VAF is tough to interpret, especially with different levels of tumor and how much they're shedding.They've told us not to rely on that in the least bit. We've asked them, “Could you give us a clear list of genes? And could you make these reports a little more obvious for the average person?” We are continuing to work with them to see if they can give us any information. And they do have some pretty sophisticated algorithms to try to tease out what is in the germline, but again, it's not going on the report, so it's of limited value unless you dig a little deeper.Yes, we've certainly seen them coming, and these are mostly a lot of lungs and noncanonical tumors. When they see a BRCA, people just don't know what to do, so it winds up in genetics. That's OK.

Mark Robson, MD: Everything does. I think one of the important take-home points for the audience is that these ctDNA assays are not germline assays. I think we all have been taking that as a given, but I think we should underline that, just like not finding a BRCA mutation on tumor testing doesn't necessarily mean there's not one present. It's even less of an exclusion if you're just looking at a ctDNA assay. You can't assume that it gives you the same kind of information. I will also kind of tease you. We've gone through actually a substantial proportion of our impact database. Among 20,000 or 25,000, we did find a very small number of mosaics. P53, perhaps not surprisingly, is the most common. In situations where you really would not expect it, you see “de novo” P53 mutations all the time. But how many of the “de novo” P53s are actually from parents who were mosaic and then transmitted it?We submitted that for publication somewhere. I don't remember.

Dana Farengo Clark, MS, LCGC: We're looking into those as well, all of our TP53 mosaics. That's become challenging. The germline labs are not doing a great job, because, as Dr Henson said, anything with a low frequency (under 30%) they call “mosaic.” I think it's between 10% to 30% that they'll call “likely mosaic.” Some P53s get special interpretations, so when someone who's even in genetics gets one of these reports, I think it's very confusing. I think the world of mosaicism is becoming interesting.Just trying to figure out whether people have Li-Fraumeni or not is becoming more challenging.

Mark Robson, MD: This is a little bit off topic, but just pragmatically, you find these alterations that somebody says, “Maybe this is a mosaic.” How do you deal with that? Because we can just give them biopsies or whatever, other tissue sampling to try to confirm it, but what do you think the most practical way of dealing with that is with regard to the family? You have an idea? How do you deal with that? You can't resolve it, necessarily.

John Henson, MD: The best way to get started with it is to do triad testing. If you test mom and dad and you can't find it, although there are certainly other explanations for why you can't find it, it is very likely that the individual has a de novo mutation. Of course, the only place that you see mosaicism is in de novo, because mosaicism is a postzygotic phenomenon, maybe at the 8-cell stage or the 100-and-whatever cell stage. But if you want to make the mosaic argument, you should not be able to find that mutation in the parents. That’s the first step. If you find it there, then that would be difficult. You wouldn't find it. That would break some fairly significant rules. I think the idea of triad testing is where you go first if you're trying to make an argument about mosaicism. Then you can look at white blood cells and try to figure it out. RNA can be helpful to see what kind of transcripts are coming off the genes and whatnot, but it's complicated.

Mark Robson, MD: We've taken a very pragmatic approach. If we can't exclude mosaicism, we test the kids.

Dana Farengo Clark, MS, LCGC: That's what we do. Test the kids.

Mark Robson, MD: If you can't exclude it, test the kids.

Dana Farengo Clark, MS, LCGC: That's what we do.

Mark Robson, MD: This is pragmatically easy to do, and that way, if the kids are negative, it doesn't matter whether they're mosaic or not.

Dana Farengo Clark, MS, LCGC: We have a very low uptake on skin punches.

Mark Robson, MD: Yes, we don't do too bad. I just wanted to show a couple more. This was the other abstract that was given to us to look at, which is this idea of using ctDNA to guide germline testing. This was interesting. A number of samples were analyzed by Guardant. Those that had either a high allele frequency or those with BRCA1 or -2 were then referred out for testing. There were a number of patients who actually did have germline mutations with a high MAF [mass air flow], and they turned out to be pathogenic. Some of them were actually of unknown significance. So there also are a number of other BRCA2s at low frequency. Not all of these actually turned out to be germline. You can see CHIP in just about everything, actually, and you can certainly see somatic alterations in just about everything. But I agree with what you said, Dr Hanson. If you have a high allele frequency or if it's in a gene that's particularly meaningful to you, then that's certainly an area to chase dedicated germline testing.

Transcript Edited for Clarity