Read the Full Video Transcript

Ashish Kamat: Hello, everybody, and welcome to UroToday's Bladder Cancer Center of Excellence. I'm Ashish Kamat, Professor of Urologic Oncology at MD Anderson Cancer Center, and it's a pleasure to welcome to the forum once again someone who's been here before, Professor Alex Wyatt from the Vancouver Prostate Centre. Alex, thanks so much for joining us and taking the time to present your talk that you gave at the plenary at ASCO GU, extremely well-received. You took a complicated topic, ctDNA detection characterization, and distilled it into a short topic.

Before I have you talk science, could you share with us and the audience what got you really interested in this field?

Alexander Wyatt: Yes. First of all, thank you very much, Ashish, for having me today and for the kind words about the talk. In terms of what got me into the field, I actually started in genetics—so human genetics. But what I found the most interesting was studying individual mutations.

Of course, when you're studying genetic disease, there's often only a small number of mutations. And so it was an easy transition to move to cancer, where the complexity is 1,000 times more, and there's loads and loads of alterations.

And I think when I first got into cancer, we were really limited by availability of tissue. We wanted to study the disease, but particularly in prostate cancer, which I started working on, there just wasn't enough tissue to study. So at that time, everybody was turning to blood as an alternative source for tumor material. And that's kind of—I guess I got in at the right time in many regards.

Ashish Kamat: Yeah. So as a field—because I do mainly bladder cancer—I think we're fortunate that we were able to get you into the bladder world because it's been great working with you and collaborating and just looking at all the advances that you made. So with that, let me hand the stage over to you and be quiet and listen to what you have to say.

Alexander Wyatt: Thanks very much, Ashish. So yeah, so I'm going to go over the talk I gave at ASCO, which is kind of touching on some of the technology and the applications of the technology around ctDNA. And I think it's nice to reflect back a little bit and, say, point out that this is not a new concept to go after the liquid biopsy in cancer.

And there was a pathologist working—actually, a resident in Melbourne in Australia—150 or so years ago that first observed circulating tumor cells (through a component of the liquid biopsy) in one of his cadavers. And it's really taken all of that time before we can actually start to unlock the potential of this liquid biopsy.

So I'm not actually talking about circulating tumor cells today. It’s really focusing on ctDNA. And that's just one analyte of a total liquid biopsy. On the right-hand side, you can see all different types of material that's present in blood or urine or other samples. But it's really the cell-free DNA that's probably become the most established analyte.

And so what does cell-free DNA come from? So it's apoptosing cells, predominantly, shedding cell-free DNA into the blood. So they're dying, and their contents are released into the bloodstream. And most of your normal cell-free DNA is coming from your hematopoietic lineage.

The median fragment length is quite small. So these are small pieces of DNA. And the median is 167 base pairs, which is exactly the interval with which the apoptotic machinery of your cell cuts. And so, as such, the fragmentation patterns that come from within cell-free DNA are biased by where the nucleosomes were positioned in cells.

And on the right-hand side of this slide, I'm just indicating that the maternal cell-free DNA placental testing—non-invasive placental testing—was light years ahead of where cancer ctDNA testing was until more recently. And that really pioneered the field of noninvasive, blood-based, cell-free DNA testing.

So your average person has about 5 to 10 nanograms of DNA in one mL of plasma. That may seem like a bit of an abstract amount, but it corresponds to about 1,500 copies of the genome. So that kind of gives you an idea of how many genomes you can survey if you have a tube of blood.

So what about the difference with ctDNA? So in a person with cancer, some of your cell-free DNA is tumor-derived. We call this the ctDNA fraction. It's a little bit smaller than cell-free DNA. But as you can see in the bottom right corner here, the distributions very much overlap between cell-free DNA and ctDNA. So you can't just use ctDNA length to distinguish it. So instead, we typically look for things like mutations that are somatic, or structural rearrangements, or even methylation marks that shouldn't be there on normal cell-free DNA, and we use those to find the ctDNA.

Interestingly, and somewhat counterintuitively (because it comes from dying cells), the amount of ctDNA is actually related to the volume of active and proliferative cancer. And this is rapidly suppressed by effective anti-cancer therapies. So ctDNA goes down when you start effective treatment.

But it's also influenced by cancer subtype. And there are some cancers that shed really high levels of ctDNA. Fortunately, urothelial cancer is an example of that. And there are others, like kidney cancer, that shed much lower amounts of ctDNA. And even within a cancer, there are different subtypes of cancers that will express much higher levels. And so small cell type neuroendocrine cancers tend to shed some of the highest amounts of ctDNA across cancers.

I think it's important to remember that when we're looking at a tube of blood to analyze ctDNA, it's not just the normal cell-free DNA in there from your blood lineage and the ctDNA. There are other potential confounders in there, too. And probably the one that people have heard about the most is clonal hematopoiesis, or CHIP. And this is mutations in your blood lineage, and they're pretty common in an elderly population. They come with a slightly elevated risk of MDS, or even leukemias. But that risk is relatively modest, and CHIP is actually quite pervasive in an elderly population.

So what we tend to recommend is that when you're sequencing the cell-free DNA, that you also profile the white blood cell DNA at the same time because then you can detect all your CHIP variants and filter them from your ctDNA calls.

So how can ctDNA tests help manage cancer? Well, probably many are familiar with there being a continuum of disease and ctDNA tests being able to help at different parts of this, all the way through from cancer screening through to the identification of resistance mechanisms and the characterization of biology. But I think what's critical is that you can't apply the same type of test to all of these different questions.

So, first of all, to highlight, in the early setting, there have been papers suggesting that we could actually even improve early diagnostics—so actually identifying cancer prior to when it currently clinically presents by finding ctDNA in the blood of, say, at-risk individuals. We can also use ctDNA detection to infer minimal residual disease. And Lars Dyrskjøt's pioneered much of this work in urothelial cancer.

In slightly later disease, we can monitor how much ctDNA there is in the blood as a surrogate for tumor burden, and this can help tell you about response to therapy, for example. And there's ongoing efforts by the RECIST working group and others around the world trying to understand the extent to which ctDNA dynamics could even be a kind of intermediate or early endpoint for our clinical trials.

And once you have metastatic disease, the burden of ctDNA seems to be quite strongly prognostic across multiple cancer types, not just urothelial cancer. But we're seeing this in multiple studies in urothelial cancer.

And then in the very late setting, we can actually use ctDNA to find genomic alterations that may be predictive of response to therapeutics. And HER2 amplification or FGFR3 mutations are probably sentinel examples of that. And then, of course, one of the key advantages of blood sampling is that you can do serial collections and actually start to look for things like resistance mechanisms arising in real time.

Now, as I kind of alluded to a couple of slides ago, a key point I'd like to get across is that you must tailor the choice of your test according to what your clinical question is. And I tend to think about two general categories of ctDNA tests, which I'm showing on this slide.

If we imagine that we have a fire, your smoke detector, on the left-hand side, must be really sensitive, but it doesn't need to tell you about what the cause of the fire is. And actually, we can use machines to do a really good job of detection.

And actually, this is quite a good analogy for ctDNA detection, where actually machine-learning approaches and algorithmic computational approaches are now highly sensitive for detection of ctDNA—so finding any molecules in the blood that come from a cancer.

However, if we actually want to understand, in this context, the cause of the fire—what are its properties, how would we target it—now you need to characterize that problem. So in the context of ctDNA, now you know the cancer is present. But maybe you want to find out, is there an FGFR3 mutation? Is there some sort of lineage plasticity going on? And I would say this is a little bit more requiring human expertise at the moment and certainly is more complex, perhaps, than the ctDNA detection problem.

So if we map this back onto the kind of continuum of disease, we can see that ctDNA detection is more important in earlier stages of disease, and characterization is more important once you have metastatic disease. Because obviously, you already know the person has cancer, and perhaps just detecting ctDNA isn't going to add that much more beyond your current clinical workup.

So let's start with ctDNA detection. And I think many are most familiar in urothelial cancer with the concept of this contributing to our understanding of minimal residual disease. Within this category, there are two types of tests. And on the left-hand side here is probably the historical gold standard—what we call tumor-informed tests.

So what you do here is you sequence a piece of tumor tissue to find personal mutations that are specific to that cancer. And then you take your blood sample, and you look with a powerful magnifying glass in that blood sample for any indication that those mutations are present. So the big advantage of this is it's very specific to that person's cancer. It's also quite established, so people are familiar with the technology.

But a downside is that it takes some additional time because you have to first profile the cancer tissue, which may not always actually be accessible, before you can profile the blood sample. So because of that time factor, there's been kind of a push towards so-called tumor-naive tests, which are on the right-hand side here.

And instead of using a tumor tissue, you kind of cast the net a bit more broadly and try to find any feature of ctDNA that's present. That could be mutations. It could be methylation marks and so forth.

So this is quicker. You could also use it for screening patients when you don't have tissue samples yet. But, as you can imagine, because you're no longer using personalized markers, the specificity is a little bit lower for detection of cancer.

And to use another analogy, I like to think about a pirate looking for a treasure, and your tumor-informed test—he has a treasure map. So we can really focus where he's looking. Whereas your tumor-naive test, you don't have a map. So those tests need to cover more ground. They need to look at more aspects of the genome. But I think it's important to recognize that technology has made this a lot more simple in recent years. So in this analogy, our pirate actually has ground-penetrating radar to help cover that additional ground.

So technology is certainly helping both the tumor-informed and tumor-naive tests become more sensitive. And generally, this is about using more features. So the older tumor-informed tests use a handful of mutations. The newer ones are using more mutations and fragment features, profiling more blood to have a better probability of finding ctDNA. And similarly, those tumor-naive tests are also incorporating methylation marks, fragment features, and even scanning the entire genome for mutations. So I really expect to continue to see the sensitivity for these tests improve over time.

So what about characterization of ctDNA? Firstly, well, why would you do this? We sort of touched on this a bit earlier. It's typically applied to advanced cancers—metastatic disease in particular. You may want to find an alteration which is associated with response to an approved therapy, like an FGFR3 mutation. You may want to actually try to infer cell phenotype, such as the transcriptional subtype of your cancer. You may want to understand resistance mechanisms or even define new tumor biology that could be targetable.

So, first of all, we may look for genomic alterations. And this is probably, I think, what people are most familiar with in terms of characterizing ctDNA. And typically, what's available at the moment commercially are these targeted sequencing panels that look across somewhere between 100 and 500 genes. They tend to be much less sensitive than those detection tests. But that's actually OK because in the advanced setting you have higher amounts of ctDNA. So it's a bit easier to see those mutations.

I want to point out three variables that I think everybody should care about when you're doing genotyping of ctDNA—so you're looking for, say, an FGFR3 fusion or mutation. The first is ctDNA fraction. So this is how much ctDNA you have in your sample.

And so when you have low amounts of ctDNA, you actually can't discriminate between a positive and a negative result. Your result is actually inconclusive because you basically did an expensive germline test. And so this is really important that we understand when the cancer was actually tested.

In prostate cancer, we've been able to develop a kind of predictive calculator that can take clinical factors and tell us which patients are most likely to have high ctDNA if we're kind of thinking about decision of whether we should try to do a fresh biopsy for tissue or profile ctDNA. And I would argue this is probably possible for metastatic urothelial cancer if we can, as a group, gather enough samples together.

Secondly, I think we should all care about alterations beyond mutations. So many tests just report mutations to us. But there are also deletions of DNA or structural rearrangements that can occur that may be just as relevant as a mutation. So I think it's really important that we be aware of what our tests don't report to us. What does a negative result actually mean?

And lastly, clonal hematopoiesis. We touched on this earlier. But clonal hematopoiesis typically affects leukemia genes. But it can also fall in some bladder cancer–relevant genes, such as TP53 and ATM. And if you just profile the cell-free DNA, even with improved bioinformatic filtering, there's a chance that you may not be able to resolve some of these mutations—whether they're ctDNA or CH in origin.

And just to highlight some unpublished data from my team looking at around 300 patients with metastatic urothelial cancer or renal cell carcinoma, when we profile the white blood cell portion and the ctDNA portion, we see that mutations in ATM and TP53 are pretty equally distributed across both the CT compartment and the CH compartment. So I think there's a risk if you're not profiling the white blood cell fraction that you may erroneously believe these mutations are coming from your cancer.

So that's kind of ctDNA genotyping in the late-stage setting, but what about the epigenome? And this is really the emerging area where technology is causing breakthroughs, I think, in the late setting because it seems that we can use methylation sequencing, nucleosome fragmentomics, and even ChIP-seq for histone modifications to get at epigenomic features of ctDNA from the blood.

And this works because those ctDNA fragments still carry those epigenomic features—those marks that were present in the tumor cells. And so we can measure them in much the same way that we could measure them from a tumor tissue sample.

Without going into any detail about any of these, what I would say is that the early data seems to suggest that they're pretty good for understanding the cell of origin for your cell-free DNA. So by that, I mean you could tell whether you have a bladder cancer versus a prostate cancer versus a kidney cancer by looking at these epigenomic marks. And probably beyond this, you may even be able to infer the actual subtype.

I think in prostate, for example, you can clearly tell the difference between an adenocarcinoma and a small cell carcinoma. I think one opportunity in urothelial carcinoma is that maybe we can tell the difference between luminal or basal and neuroendocrine features using any of these approaches.

There's even opportunities to understand the normal cell-free DNA signal to actually have a better picture of patient biology. And I'll highlight one anecdote from a paper we published recently where we used ChIP-seq directly from plasma, and we were able to observe the epigenomic footprints of metastasis-induced normal tissue destruction.

So, in this particular case, we were seeing cell-free DNA from the destruction of colorectal cells by a prostate invasion into the colorectal space. And so we were seeing that epigenomic footprint of those GI tract cells in the blood at the same time as we were seeing ctDNA.

I think this technology is very much in what I would call the pioneer phase. So it's immature relative to those genotyping tests. And I think it has a clear role as a biological probe to help us understand disease. But I think the jury's still out on the clinical utility. There's not yet been any head-to-head comparisons between those different types of tests. So we don't know which is maybe better than the other or whether they're complementary. I think, importantly, the sensitivity and lower limits of detection for each of those technologies remain pretty unknown. And I believe that they're going to vary by what you're looking for in the blood.

If you're looking for something that's a really extreme difference, perhaps you can do that at low levels of ctDNA. But if you're looking for subtle differences in, say, a single pathway expression level, you're probably going to need quite high levels of tumor DNA in the blood to have good accuracy there.

There's still some patient biology we need to understand—what influences ctDNA fraction from day to day, from hour to hour? What influences the normal cell-free DNA that's shed into the blood? We still don't fully understand that.

And the final thing I'll touch on before I close is just to mention urine DNA because this is an exploding area for diagnostics in urothelial carcinoma. And I think there's clear promise to improve UC diagnostics with urine.

A slight difference between plasma cell-free DNA is that while urine is, again, a mixture of cell-free DNA and cellular DNA, in the context of urothelial carcinoma, we actually really care about those urothelial cells. And that cellular DNA is very relevant for understanding urothelial cancer. So it's perfectly viable not just to study the cell-free DNA, but also the cellular DNA from urine.

Having said that, I think many of the lessons that I've been going through today with respect to plasma are also relevant for urine DNA. Those same technologies are also unlocking the potential of the urine liquid biomarkers.

I would say the specificity for mutations is potentially a little bit less clear because we know we have field cancerization within the so-called normal urothelium. So I think we have a lot of work to do to understand how to tell the difference between something that's in, say, a somatic field versus something that's bona fide coming from a cancer.

And I think as well, the urine field is a little bit behind plasma, where we have very clear standardized protocols for how much blood and how to collect blood, what preservatives to use in urine. I think we need to still do some standardization here before we can reach some consensus as a field.

So just to conclude, I think the key message I would get across today, the one thing I would want everybody to take, is that one size does not fit all when you're selecting a ctDNA test. You really have to tailor your selection of technology to what clinical question you're trying to address. And at a big, higher level, that's whether you want to detect cancer or you want to characterize it.

I would say that we still have more work to do to optimize both sensitivity and specificity, particularly for a tumor-naive test. But it's really exciting to see this new wave of epigenomic technology—so your methylation profile and your fragmentomics. The promise of that for boosting sensitivity and actually helping us to understand the biology of cancer cells is super promising.

My final sort of words would be choose wisely and really be aware of the limitations of technology because I think it's only by embracing and understanding those limitations that we actually unlock the potential of testing.

So with that, I'll say thank you very much to UroToday and to Ashish for having me on. And happy, if there's any time left, to talk about some of these points.

Ashish Kamat: Thanks so much, Alex. I mean, there is so much that we could chat about. But that being said, I mean, you covered everything so well that for the purpose of education and didactics, I think this was an excellent talk. And I learned a lot, too. So I want to close this session and say, thank you very much.

Alexander Wyatt: Thank you very much for having me.