Welcome to today's webinar, Liquid Biopsy in Hepatocellular Carcinoma, Applications in Surveillance and Prediction of Treatment Response. Today's webinar is presented by AASLD's Hepatobiliary Neoplasia Special Interest Group. Today's webinar presenters are Drs. John Kissel, Louis Roberts, Augusto Villanueva, and Dr. Ju Dong Yang. Dr. Villanueva will also serve as the moderator. Today's webinar will be available on demand on AASLD's online learning platform, Liver Learning. Be sure to connect with us on the following social media platforms. Also tune into our YouTube channel to view our videos. Today's a great time to join AASLD if you're not a member. As a member, you can join as many special interest groups as you would like. Listed are a few other member benefits. Check out the new AASLD Journals app. Access to all four AASLD journals are in one place. Visit aasld.org slash publications for information. AASLD has many upcoming meetings covering a variety of topics. Visit aasld.org slash calendar for a full list of our events. The Liver Meeting 2019 is November 8th through 12th in Boston, Massachusetts. Please visit our website for more details. DDW is May 2nd through 5th, 2020 in Chicago, Illinois. The abstract submitter is now open and closes December 1st. Visit ddw.org slash abstracts to learn more. You can help support the future leaders in hepatology by giving a donation to the AASLD Foundation. Your donation in any amount is tax deductible. Visit aasldfoundation.org slash donate to learn about ways to give back. Please feel free to submit questions throughout the presentation using the Q&A box at the top or bottom of your screen. All questions will be answered during the Q&A portion of the webinar. Now I'll turn it over to our moderator, Dr. Villanueva. Good morning, everyone, and welcome to the AASLD webinar on liquid biopsy in HCC. We'll have three presentations today. The first one, I will give an overview of what is liquid biopsy, the different technologies that are available, and frame a little bit the discussion in terms of what are the potential applications of liquid biopsy. Then we'll have Dr. Kiesel from Mayo Clinic, who will discuss specifically methylation analysis of circulating tumor DNA in the context of early detection. And finally, Dr. Yang from Cedars-Sinai that will discuss prospects on the use of circulating tumor cells for prognostication and eventually prediction of therapy in HCC. So these are my disclosures, and this is the overview of the first part of the webinar. So I'll discuss a couple of unmet needs in clinical management of HCC, give you an overview of what is liquid biopsy and the potential applications, and then we'll move ahead to discuss some potential implementations in the context of early detection, how we can trace tumor burden, and eventually detect minimal residual disease. I would like to start just by giving an overview of where do we stand in terms of disease burden related to HCC. You have data from GlobalCAN 2018, estimated number of total deaths and estimated number of new cases for different tumor types. The size of the dot represents the ratio between deaths and incident cases, which is kind of a surrogate of the lethality of the tumor. So number one at the top, the most frequent malignancy is lung cancer, and it's one of the more lethal. Then you have here liver cancer, which is the fourth in terms of mortality, just right behind the stomach. So we're dealing with, as you know, a very deadly disease where we have limited treatment options for patients diagnosed outside early stages, and that's why it's so critical to identify these patients when we can apply potentially curative therapies. There is one thing that is particular in HCC is that we have a very, I would say, decent understanding of the natural history of the disease. So it's very rare to develop HCC in patients without chronic liver disease, and here we have an example of the natural history of NAFLD-related HCC. As you know, the risk of HCC in patients with cirrhosis, regardless of the tolerances between 2% and 3% per year, in NASH is a little bit lower when it's compared with hepatitis C. And also we know that, for example, in NASH patients you can develop HCC even in patients that have not developed cirrhosis. The prevalence of that is very low and it's still unclear, but the point here is that we know the natural history so we can enroll the patients at high risk in surveillance programs so we can detect tumors when they're very small, when they're at early stages, so we can apply curative therapies. The current standard in terms of surveillance for HCC is to conduct ultrasound and plus minus alpha-fetoprotein, depending on the guidelines, whether it's American or European, every six months. So MidSingel reviewed last year, which is the performance of ultrasound plus minus AFP for the early detection of HCC, meaning detection when tumors are at early stages, and the sensitivity was 63%, which means that around 40% of the patients with early tumors are not detected at early stages. To give you some perspective or where do we stand in terms of biomarkers, of plasma biomarkers for early detection, alpha-fetoprotein has been associated with HCC. The first report that I could find was a Lancet paper in 1970, so almost 50 years ago and nothing has significantly changed in terms of blood biomarkers for early HCC detection, and hopefully liquid biopsy will change that in the near future. HCC has a number of singularities, and I just want to focus in some molecular features that are quite unique in this malignancy. So this study conducted a thorough analysis of gene expression data from 1,000 tumor samples that are deposited in the TCGA, which as you know is one of the largest genomic initiatives to profile human cancer ever conducted. When authors cluster the samples based on their similarities, this is a PCA plot that essentially shows the relation of the different samples. Each dot represents a tumor sample and their color based on the tumor type. So these are the different tumor types that were analyzed. What you see is that in principal component one, what you're capturing is essentially the different tumor types, which to some extent makes some sense. But principal component two is basically capturing the difference between these samples and the rest. Well, actually these samples are the HCC. When you group all the samples together for the different tumor types, you see that HCC is an outlier here, relatively far away from the other tumor types, which suggests that when we look at the gene expression, there's something unique in HCC compared to other tumor types. There's another singularity in HCC. So the people from Memorial Sloan-Kettering analyzed, again, a large number of samples from different tumor types, 10,000, and they were specifically looking for draggable mutations. So they selected a panel of 500 frequently mutated genes. And then they analyzed these 10,000 different tumor types with these panel of mutations. What they did is that they rank the mutation based on the levels of draggability. In other words, what are the chances that a given mutation could be used as a biomarker to predict response to a specific therapy? Level one was when the biomarker had a FDA approval for a specific FDA approved drug in that same indication to give you scope of the type of results that they reported. As you can see, around 40% of all the samples had at least one draggable mutation. What is interesting and what is relevant in HCC is when they rank the different tumor types based on the number of draggable mutations. And you have here, so G is the tumor with the highest proportion of draggable mutations, as you know. HCC, I'm sorry for the, so the red box should be over. Hepatocellular carcinoma was second to last. So only mesothelioma had less draggable mutations than HCC, which poses a problem in terms of implementing precision medicine as, you know, despite we know which is the mutation profile of HCC, there's little that we can do in terms of predicting response to the mutations that are detected. So to summarize the first part of the clinical challenge that we have in HCC, we know that early detection tools are suboptimal and most cases are still diagnosed at intermediate or advanced stages. We know that there is a lack of biomarkers of response to select the best candidates for systemic therapies. You know, in the last three years, there has been a lot of changes in the management of HCC with now seven drugs approved by the FDA, some of the systemic agents approved, but there isn't clear which is the best sequence of combination therapies or how to sequentially give these therapies to maximize response and minimize cost and toxicity. So it would be good to select using biomarkers who would be best responder at each stage of the disease. And we have an additional problem. As you know, HCC can be confidently diagnosed with imaging techniques, which means that we don't have, compared to other malignancies, significant access to tissue for biomarker studies. And that's problematic in patients with advanced stages since we don't have biopsies. We don't, it's very, it's more difficult to conduct biomarker studies that would identify those eventually markers that could help us organize or structure the treatment based on response rather than other variables. And following as a result of that, it's very difficult to conduct molecular monitoring. I'll talk a little bit later on what's the implication of molecular monitoring, how it can be used particularly to identify mechanisms of resistance to systemic therapies. So moving to the next, to the second point, what is liquid biopsy and what are the different technologies that we can analyze? So liquid biopsy essentially describes the analysis of tumor components that are released to the bloodstream. Well, essentially to any fluid of the body. It could be saliva, it could be ascites, but most of the studies have been conducted in peripheral blood for its convenience essentially. And there are three major components that can be analyzed. The first one are the actual cells that are released to the bloodstream. Those are called circulating tumor cells or CTCs, and that's the biological substrate of distant metastasis. You can have also exosomes in the blood. Exosomes are small vesicles. There are different types, and I will not go into much detail, but we know that they're functional. So their content, mostly small RNAs, but also they can have some DNA and some proteins. They are rather sophisticated mechanism of communication between cells. And there are plenty of studies that suggest the direct oncogenic role in tumor progression, including HTC. The third component, and one that is, I would say, more developed, are or is circulating free DNA. So these are fragments of DNA that are diluted in the plasma. It's important not to think that this circulating tumor DNA comes from the CTCs because it's not the case. They're two different compartments of the liquid biopsy definition. CTDNA is diluted, whereas CTC, the DNA is intact and it's within the whole cell. Now, one of the applications, and going back to the molecular monitoring concept, is that, unlike tissue biopsy, it's very easy, very convenient to take a blood sample from a patient. You can do that sequentially. What that allows is to real-time follow or monitor, which is the clonal composition of the tumor. So here you have an example in a lung cancer patient. So you have a patient with, you know, different tumors in different locations of the lungs and the liver. And we know that tumors heterogeneous, or the different cancer cell clones that form these tumors that are depicted here in different colors. So at baseline, what you have in this patient is a predominance of the gray tumor clone. So most of the cells have gray. But when you treat the patient and you're able to specifically target that clone, what you see is a change, a shift in the composition of the tumor clones. And you have an emergence of the green clone that may not be responsive to the therapy that was initially given for the gray clone. So to trace that using biopsy, as you can see, it's quite complicated, as you can need to biopsy different tumor nodules. It's invasive, may have potential complications. But in blood, it's very easy. In the beginning, you can quantify the DNA that was released by this gray clone. You see it goes back. And then when the green tumor clone emerges, you can also trace that minimally invasive with analysis of seed DNA. So again, it's a very convenient tool to monitor how tumors evolve over time. And that dynamic perspective is a significant advantage compared to conventional tissue biopsies. Now, what do we have in blood? So when you take a blood sample and you centrifuge it, you have here the red blood cells. You have a thin white layer called buffy coat that essentially has the immune cells and the circulating tumor cells, and the majority, around 55% of that tube is plasma. And here in plasma, sorry, these arrows should come from plasma, there's where you have the circulating, the diluted DNA. The amount is between 30 and 180 nanograms per ml. The majority of the DNA that is in plasma is not from the tumor. This is an important message. So all the cells in the body in, you know, normal turnover, when they die, they release non-tumoral DNA to the blood. So the blood is a composite of DNA from different sources, and a percentage of that comes from the tumor. What percentage? Well, it's estimated that less than 10%, but it's probably even less than 2% for the majority of cases. So less than 2% of the DNA that is in the blood actually comes from the tumor. The length of the DNA that is diluted in blood is very stable. It's approximately between 140 and 170 base pairs. It's very stable. And the half-life of circulating tumor DNA is around approximately two hours. And we know that from studies of patients that were treated with resection. So you have the tumor, you take a blood sample right before the resection, and you take sequential blood samples after the resection. So after two hours, you don't detect any more ctDNA. So it's very rapidly clear from the circulation. Most of the studies that analyze circulating free DNA began from non-invasive prenatal testing. That's where the technology exploded, and now it's mainstream, and it's commonly used for the detection of chromosomal trisomy and other chromosomopathies from pregnant women. There's a more recent iteration of this approach that tells you and shows you the power of this technology. So similar approach, but in this case, instead of looking for trisomy of chromosome 21 in a pregnant woman, because the concept is the same, so the genetic makeup of the fetus is different from the mother. So fragments of the DNA from the fetus can be accessed by a blood sample conducted on the mother. So in this case, they analyzed 30 genes for monogenic disease, which is, let's say, a more specific and granular analysis of alterations in cell-free DNA. And to, you know, summarize, make a long story short, this is the results of this analysis that included 233 pregnant women, 20 true positive, and 127 true negative, meaning that a monogenic disease of the fetus was detected by analyzing the blood of the mother, including osteogenesis imperfecta, chondrodysplasia, and other monogenic diseases. So you have here a proof about the, you know, accuracy of this technology to detect, in this case, mutations in monogenic diseases. Implementation in clinical practice of analysis of ctDNA for mutation in cancer. We have some studies, not in HCC, there's some coming out, but not yet published, in terms of how we can use this information to select, for example, for clinical trials. In this study, they analyzed around 100 patients. They took the blood, they analyzed the tissue, and for 75 of the patients, 75% of the patients, there was a concordance between the findings in the tissue and the blood. And out of the 100 patients analyzed, this is the first part of a very long prospective study, 11 patients could be matched to a clinical trial. So for that, you need the patient to have a mutation that is druggable and that is detected in tissue and blood. So it's still a bottleneck on how we can apply this in clinical practice, but as you can see, out of 100, 11 enroll in a trial for partial response in seven stable diseases. So what about edit detection in HCC? I'm going to give you a summary of where do we stand in terms of how we can implement this for edit detection. There's estimates that around 600,000, there are 600,000 carotid cases in the United States. If we assume that between 2% and 3% are at risk, annual risk of developing HCC, we have 18,000 new cases of HCC every year. Ultrasound and alpha-fetoprotein detect 63%, so we have approximately 6,000 patients not detected using conventional surveillance techniques. Well, this is the ideal scenario, but the reality is that out of the 600,000 patients at risk, or out of the 18 that would develop ATC on a given year, not all of them are under surveillance programs. Actually, a minority, around 25%, are enrolled in surveillance. So if we apply the performance of the surveillance techniques in this set, we have only 15% of the total amount of potential patients developing ATC in a given year being detected with ultrasound and alpha-fetoprotein, which leaves us with 85% either not detected because the tools are not good or because the patient is not enrolled. So these are the two main barriers to implement ATC surveillance and to increase early detection rates. So we need a tool that is accurate, that is reproducible, that is easy to implement, and ideally, the cost should be reasonable. I just want to point out the fact that the importance of this early detection tool to be reproducible, meaning that it's not dependent on the operator, which is one of the major problems of ultrasound. If you see the data of ultrasound performance in early detection, you have huge differences depending on who or where the procedure is conducted. Now in terms of using this approach to trace, to track clonal composition, this is, in my view, one of the most comprehensive analyses of such, or aimed at following or tracing clonal composition in cancers, in Lyme cancer, again. Authors, they analyzed 100 patients. They did multi-regional sampling in the tissues, whole exome sequencing. They could generate phylogenetic trees, or in other words, they could estimate the clonal composition in the tissue, and based on that data, they design specific assays to detect those mutations in blood of the same patients, and they follow those patients over time. It's a huge effort. What they saw is first that there's a direct correlation between the amount of mutant DNA from the tumor detected in the blood and the size of the tumor, and we're talking about binary fraction in the blood of 0.1, and tumors with a volume of 10 square centimeters, so the diameter of this tumor volume is around 1.3 centimeters, so we're talking about very tiny little tumors, and you're able to detect already CCDNA, and as expected, what shows here is a sequential sampling of this given patient is that the detection or the identification of mutant DNA in the blood precede the actual identification of recurrence using conventional radiology techniques, so it seems that in addition to being more convenient, it's more accurate for the detection of recurrence. So the advantage of this approach in Lyme cancer is that we have tissue to analyze, but there is a specific mutation in Lyme cancer that is trackable, a thing that we don't have in HCC, and that's how the major bottlenecks for implementing these for HCC. We've done some preliminary studies to see if we can detect mutations in patients with HCC. This is a pilot study from eight HCC patients at early stages for which we have blood and tissue. We show that when we analyze the 60 most frequently mutated genes in HCC, we can detect around 70% of the mutations in blood that are also present in tissue, and we validate that using other approaches, specifically digital droplet PCR. We have some technological improvements, such as the incorporation of molecular barcodes and other, let's say, technical modifications that has increased that detection rate to now around 85%, and that will be presented by Dr. von Felden in the upcoming liver meeting. We have the details here of the poster where he presents this data. And finally, moving a little bit beyond circulating tumor DNA, there's also data on exosomes. So we've done also this study analyzing circulating extracellular vesicles, exosomes in patients with HCC. We confirmed that we can detect these small vesicles that you can see here. The size is around 120 nanometers, and we know that the content is mostly RNA. By doing RNA sequencing, we can generate specific signatures that truly differentiate HCC patients from controls in this pilot study of 15 patients. And the other interesting thing is that when you compare the analysis of exosomes and plasma from the same patients, in other words, do we need to go through this, I would say, rather complex process of ultracentrifugation to isolate the exosomes? And the answer is yes, because there are some microRNAs in this case that are present in the exosomes that are not detected in plasma. So to summarize this first talk, I would say that it's very clear that tumor cells release components to the bloodstream, mostly DNA, RNA, and CTCs, and this can be isolated and analyzed and provide or help us understand which is the molecular profile of tumors by using just blood. The analysis of CTDNA allows detecting mutations in patients at early stage. This is applicable in different malignancies and also in HCC. That secreted exosomes contain RNA that can be also analyzed and help discriminate between HCC patients and controls, but I would say that this technology is less developed than CTDNA for early detection. And I think that, with any question, the good bias has emerged as a promising and very convenient tool for biomarker development in HCC, whether it may have a role for early HCC detection is still to be established, but certainly preliminary data is very promising. And I just want to thank the people in my lab, because I presented some data generated by them, particular Johan Wolfred and Teresa Garcia-Lazama. And with that, I will introduce Dr. Kiesel, who will discuss other aspects of CTD analysis for early HCC detection. John? Augusto, thank you very much for that very comprehensive overview of liquid biopsy and HCC. I'd like to focus a little bit more now on the early detection application and our work in methylated DNA as a component of circulating tumor DNA for that purpose. So, by way of disclosures, I do work closely with a company that is involved in this space. And as Augusto mentioned, circulating tumor DNA is one of several tumor-specific or organ-specific components that can work their way into peripheral circulation. And the liver is actually probably one of the most ideal organs for the use of CTDNA for a variety of disease monitoring applications because of the very high cardiac output that runs through the liver. And even normal liver DNA is a fairly abundant component of the total amount of circulating DNA that is free in plasma. Lymphocytes are the other major source. In this study that was done in the non-invasive perinatal testing laboratory in Chinese University of Hong Kong, they were able to show that a variety of markers that indicated that circulating DNA was of liver origin. They were able to show that patients who had hepatocellular carcinoma had a significant increase just in the total amount of circulating DNA that was coming from the liver. So, if you're going to target DNA, we've heard about a variety of different potentially targetable media within liquid biopsy. And we've also seen some of the impressive preliminary data looking at DNA mutations for both early detection or for therapeutic monitoring. What would the ideal target biomarker be and which component of DNA should we be looking for? We could broadly lump that into genetic categories such as mutations, large pieces of DNA like chromosomal DNA. We could also look at epigenetic components like nucleosomal DNA wrapped around histones or methylated DNA where the sequence is preserved but there's a covalent modification to cytosine residues. And because of the fact that, as Augusto mentioned, the circulating tumor DNA is highly fragmented and maybe only generally in the region of about 150 base pairs in length, that really gets us down to changes that we can measure related to the tumor at the single nucleotide or single strand level which probably brings us down to DNA mutations and DNA methylation events. As was highlighted, there's a fair amount of heterogeneity in the mutation profile of HCC and we really don't have the same intensity of hotspot mutations that were highlighted in the liquid biopsy of lung cancer in the prior talk. Some of our most abundant and commonly mutated genes in HCC are found at relatively low frequency compared with some other tumor types and so you see rapidly get down into single digits of abundance after the first three candidates. Additionally, depending on the target that you're interested in, there may be pathogenic mutations across a fairly wide landscape of real estate for a given mutation. P53, which is the most common mutation in HCC, is particularly problematic because the mutations that can occur on that gene with functional consequences occur over a fairly wide range of almost 300 base pairs across five exons and so as you're thinking about trying to surveil a population of patients at risk that's in the hundreds of thousands, it becomes very difficult to develop allele-specific assays for this many mutations and so you're kind of really looking at sequencing in order to do that which is still fairly expensive and fairly difficult to automate for a high throughput assay. Again, DNA methylation is a little bit different and the sequence of the DNA is preserved but typically in areas with dense cytosine content, these are covalently modified by the addition of a methyl group and the canonical mechanism for that is that when the promoter upstream of a gene is methylated, it blocks transcription and that often silences a gene that may have a biologically protective mechanism or tumor suppressor role for the cell. Another important attribute of DNA methylation is that the methylation events are fairly common in particular tumor types and so in this example you can see that just the first five or so candidates of methylated genes are densely methylated across almost all of the resection specimens that were included in this tissue-based study. There are a few tumors that did not get picked up but a combination of just five of these markers was more than 95% sensitive and specific in this tissue-based study and so when we say broadly informative, we mean that relatively few markers are required to represent events occurring across a fairly large number of tumors. So what about the functional roles of some of these? I don't have mechanistic data available for the majority of these candidates but in work that we did in collaboration with Lewis Roberts Lab, we were able to establish that this particular gene of interest BMP3 which is included in the multi-target stool DNA test is also densely methylated in cholangiocarcinoma cell lines but not in a normal cholangiocyte cell line and that methylation results in a sharp decrease in the expression of that gene BMP3 which can be reversed with the use of demethylating and deacetylating agents and when the expression of that gene is restored in cell lines, there is an increase in apoptosis and a reduction in cell viability and suppression of tumor growth. So these are biologically meaningful targets even though they may or may not yet be druggable. So when we set about to look at DNA methylation comprehensively in hepatocellular carcinoma, we used a next generation sequencing technique, in this case reduced representation bisulfite sequencing and it was important for us to include tumors from resection specimens that represented a fairly broad array of underlying liver diseases. We're probably biased towards patients without a great degree of underlying cirrhosis because their tumors were resectable and we also needed to look at ideal control tissues which would be not only normal or cirrhotic liver but also white blood cells which are again the major source of background methylation in circulating tumor DNA and we wanted to then use some statistical techniques and filters to get this list of candidates down to a fairly manageable number of targets for a clinical grade assay. After doing that, running through the tissue data that I showed earlier in the heat matrix and a couple of pilots, we settled on a six methylation marker panel as well as a methylated normalizing gene and tested that panel in about 100 HCC patients, 100 controls and 50 cirrhotic controls who had been followed long enough for us to be reasonably certain that they did not have underlying undiagnosed HCC. We had a couple of statistical models, one that was the best fit data and then also a cross-validated model that has run through a statistical filter several thousand times to measure the potential sensitivity and specificity and application to future data sets and that cross-validated model was substantially more predictive of case or control status than alpha-fetoprotein, the current standard of care biomarker. Using that information, we're able to assign in each case a probability of there being cancer. We can see that there are certainty of probability increases with the stage of disease but even for stage A disease, we had relatively high sensitivity. In an archival sample case control study design, these things may change as we're able to validate this in larger and more diverse data sets but this is a very strong early start for the concept of methylated DNA as an early detection biomarker for HCC. As Augusto mentioned, I think there's a really strong rationale for an operator-independent way to detect these tumors and we think we can probably do better than alpha-fetoprotein. There are a variety of people working on other proteins of interest. The GALAD model is currently under a great deal of scrutiny. There are also other laboratories, both academic and commercial, that are working on developing circulating tumor DNA biomarkers using methylated DNA. We think that this particular next-gen sequencing tool is very potent for discovery but in order to do this in an automatable and high-throughput way, at this point, we've been concentrating on a targeted assay. For those of you who will be attending the liver meeting, we're hoping to show some new results from a multinational case control study that are further selecting and refining our marker panel as well as validating the sensitivity and specificity estimates. That'll be presented by Dr. Chalasani on Sunday, November 10th and look forward to taking questions from the group in the Q&A session. I'll turn the slide deck over to Dr. Yang who will talk about circulating tumor cells in liver cancer as a monitoring approach. Good afternoon, everyone. I'm going to discuss about the circulating tumor cell in liver cancer for the next 10 minutes. I have nothing to disclose. I'm going to discuss about circulating tumor cell in hepatocellular carcinoma followed by circulating tumor cell in cholangiocarcinoma and I'll finish my talk with summary and future direction. Primary characteristic of the cancer is the uncontrolled proliferation with local invasions and introvasation of the tumor cell into the bloodstream. Few of them survive in circulation, arrest at the stern organ site, extrovasation, formation of the micrometastasis and establishing the clinically apparent metastasis. Previous study shows that about millions of the circulating tumor cells are released into the circulation per gram of the tumor each day, highlighting that the circulating tumor cell could be a potential great biomarker that reflect tumor biology by a simple pair-per-blood trial. CellSearch, the circulating tumor cell detection platform, is the only FDA-approved platform for the detection of STTC. I'll briefly introduce the technology to you. So, the blood is collected in a cell-safe tube and the blood is now placed into the autoprep system for isolation of the CTC. So, this cartoon shows the CTC and using the antibody against the epichem, which is the marker for epithelial cell origin, and this antibody is connected with the ferrofluid. So, whatever the cell that is attached to, with this antibody, can be isolated using the magnetic force. Once these cells are isolated, we use antibody against cytokeratin, and this antibody is conjugated with the trichoerythrin that gives the fluorescent signaling for identification of circulating tumor cell and enumeration, which is done using cell trick analyzer. So, all your studies on CTC in a CTA patient were done using cell search system. CTC detection in a patient who is undergoing surgical resection of the tumor. So if someone has the CTC prior to surgical resection, that predicted the post-surgical recurrence of the hepatocellular carcinoma. And also, CTC positivity was associated with vascular invasion, elevated alpha-beta protein, and advanced stage of the tumors. And also, it predicts the disease progression after taste treatment. And also, it's been associated, it was shown to be associated with the strong, the predictor of the overall survival in patient with the hepatocellular carcinoma. But limitation of the cell search system is that only about 30% of the hepatocellular carcinoma have the an applicant expression in the cell surface. Therefore, has lower sensitivity of detection using this technology. Furthermore, once the CTC enter into the bloodstream, some of these cells go through the EMT process, and thereby losing the epithelial markers. Therefore, they even have the lower sensitivity for detection of the circulating tumor cell using this technology. So people try to develop a better platform that increase the sensitivity of the CTC detection, one of which is the nano-velcro CTC detection platform. So this platform used the multi-marker panel, including the APCAM HILO glycoprotein receptors and GPC3, and this multi-marker panel this platform was shown to have our highest sensitivity, and CTC was detected in 97% of the patient with the HCC patient. Furthermore, this technology allowed the biological characterization of the CTC. This is the images from the platform. So if you look at the upper panel, this is a regular CTC that has the DAPI staining, cytokeratin staining, but no vimentin staining. Vimentin is representative of the EMT process without any CD45 staining, meaning that this is not white blood cell. So this is a regular CTC but if you have a CTC with the vimentin positivity, then this is what we call the vimentin positive CTC, and those who had the vitamin and the positivity CTC, they had the shorter overall survival and also shorter time to recurrence after local regional treatment. So more recent study and also show that we can do gene expression, the analysis from the isolated CTC for further biologic characterization of the CTC as well. So CTCs in cholangiocosmoma has not been researched much, and there are much less data in the literature, and let me show you about one of the studies that we conducted at Mayo Clinic. So we had a 88 cholangiocosmoma patient. About 20% of the patient with the cholangiocosmoma had circulating tumor cell two or higher. As you can see, demographic features were comparable between the two groups. You see that intrapartic cholangiocosmoma was more common in a CTC positivity group. CTCs were associated with the extent of the tumors, and those with the CTC two or higher were having larger tumors, multinodular tumors, and also bilovar tumors, and more likely to have the lymph node metastasis and extrapartic metastasis. Finally, CTCs were associated with a poor overall survival in cholangiocosmoma patients. For those with the CTC two or higher had a median survival of five months compared to 27 months for those whose CTC was less than two. In multivariate analysis, CTC two or higher were independently associated with a shorter overall survival in patients with the cholangiocosmoma. In conclusion, CTC seems to be useful biomarkers in liver cancer, especially in terms of the prognostication and probably for the prediction of the treatment as well. And also, it allows the biological characterization of the tumor, too. However, there is a limitation. CTC detection sensitivity is related to the tumor volume, so detection sensitivity tends to be higher for larger tumors, but the sensitivity decreases if the tumor size decreases. Therefore, there are almost no data investigating the utility of the CTC for all the detection of the tumor. And also, there are numerous CTC detection methods reported in the literatures, and most of them were a small single center case control study was conducted, so it is difficult to have a strong conclusion. I have a strong conclusion in terms of the utility of the CTC and testing liver cancer patients. So, a future direction. So, we need to have a better platform with a higher sensitivity for CTC detection with a standardized assay protocol that will enable a multi-center prospective study with a larger sample size. And we need this type of data for us to use the CTC in our routine clinical practice. Thank you. Okay. So, with this, we complete the presentations. Now, we open the webinar for questions and answers. We have already one question directed to Dr. Kiesel, and the question is that we know that dysplastic nodules might harbor mutations such as STIR promoter. What is the data or what is the knowledge on methylation changes in dysplastic nodules? So, that's an excellent question. And so, the answer is we don't yet know. We have a study that's in progress right now looking at mutation, sorry, rather methylation events in hepatic adenomas in comparison to normal liver parenchyma, cirrhotic tissue, HCC, and focal nodular hyperplasia. But I don't think it's common to have a large tissue bank of just the dysplastic nodules. And so, a lot of the times when we're seeing those in tumor specimens or in resection specimens, they may be in a field that's been affected by the tumor or in diseased liver that gave rise to the tumor. So, I suspect that we will find that many of the same methylation events that we're able to see in the primary tumors will be found in dysplastic lesions as well. We've illustrated that concept with colorectal adenomas, with dysplastic lesions of the esophagus, and with PANIN and IPMN precursors of pancreatic cancer. And there tends to also be sort of a dose-dependent effect on the severity or the grade of dysplasia as you progress towards cancer. Okay. We have another question for Dr. Yang. Does your technology for CT detection allow to customize the selection of markers? For example, can we include PD-1 or PD-L1? Oh, thank you for an excellent question. Yes, indeed. The technology allows us to evaluate the expression of the specific markers of interest. Indeed, my collaborator at UCLA specifically looked at the PD-L1 expression in CTC using the same technology, nanovelcro technology. And we actually made an oral presentation at ILCA showing data that the CTC with the PD-L1 expression 410 for prognosis in patients with the hepatocellular carcinoma. We tried to evaluate whether that predicts the responsiveness to immune checkpoint inhibitor, but we are working on that. Okay. We have another question for Dr. Kiesel. Specifically, are those methylation changes occur in hotspots of the genes? Specifically, is it able to target sections of the genes, or do you need to cover the whole CPG island? So, when we are doing sequencing-based discovery for DNA methylation events, we are studying as much of the genome for methylation as possible. The reduced representation bisulfite sequencing approach eliminates large regions of the genome, which are CPG-poor, and so gets rid of a lot of repeat sections, really enriching the genomic DNA content for CPG islands. But again, because the DNA that will end up in circulation will be highly fragmented, we are typically interested in finding areas where we have multiple contiguously methylated CPGs over a relatively short distance, typically about 100 to 150 base pairs in length. So, the CPG islands probably remain the predominant quote-unquote hotspot for DNA methylation in cancer, but we do also find gene body methylation, and we find methylation in areas of the genome which are non-coding, sometimes producing regulatory RNAs or proteins, or RNAs rather. Okay, we have another question for Dr. Kiesel. What do you think about the biosensors in terms of detecting ctDNA in small amounts? Do you think detection of methylation in ATC and biosensors will allow increased detectable ATC tumors? That's a very good question, and I'm afraid I don't know enough about biosensor technology. There's a lot of interest in developing assays for methylation of DNA in circulation that tries to avoid the use of bisulfite treatment as the first step in the assay process. So, that's a very harsh chemical reaction which converts methylated cytosines to just cytosine and unmethylated cytosines to uracil, and then the subsequent reactions, the PCR reactions, measure the differences in those post-conversion changes. But the bisulfite treatment process itself actually is very harsh chemically and destroys up to half the DNA in the sample that you're trying to analyze. So, there are a variety of people that are looking at different mechanisms like nanostring sequencing or immune precipitation or capture or restriction enzyme-based methods to avoid that bisulfite step. But I'd be happy to talk offline or receive email about a potential biosensor that might be of interest because I think, broadly speaking, there is great interest in trying to, again, avoid bisulfite conversion if possible. There's really no widely used system that we think works better, but we're always open to new ideas. We have another question of whether there is rationale of increasing the number of biopsies for HCC patients at intermediate or advanced stages in order to better characterize the tumor and predict response to treatment and also develop more evidence for performance of liquid biopsy. So, I think that there's a time where we need to make sure that potential biomarkers can change our clinical practice. And so, one of the things that would make us do more biopsies of patients for patients, not for diagnostic purposes in these stages, but to actually produce, was actually to have a biomarker that produced response to any of the new systemic therapies that are available. So, I think that there is a need to, once we have that biomarker, see how it performs when analyzing ctDNA as opposed to tissue. So, I think there's certain movements to move towards more invasive approach in terms of being more proactively biopsying patients at advanced stages to understand that we're able to take those predictive biomarkers and see how they perform in blood. We have another question for me. Do you think there is a role to play by non-invasive imaging, ctMRI, MRE, at better capturing early stage HCC and hence improve diagnostic performance of liquid biopsies? So, I think that the way I see surveillance in the future is that it's essentially removing ultrasound from the algorithm because it has some issues such as the fact that it's operator-dependent. So, if we have a blood biomarker, and I'm talking about diagnostic, this is an early detection tool, it's a surveillance tool. You will always need to do a confirmatory MRI or CT scan first to diagnose the HCC and second to stage the patient. So, I don't think liquid biopsy will eliminate imaging. I hope it would eliminate ultrasound as a surveillance tool. It will facilitate implementation as you will have only to take a blood sample, send it to a central lab, run whatever test, either methylation or mutation analysis, and you get a probability based on that of that patient having an HCC that would trigger additional tests, essentially imaging tests to confirm the diagnosis. We have another question. So, in terms of routine surveillance for HCC, how often is AFP alone used instead of ultrasound and AFP? The question is not directed towards anybody. So, I don't know. Luis, do you want to chip in? Yes, thanks. That's a great question. And I think the consensus really from all societies would be that AFP should not be used alone for diagnosis of early HCC, particularly because the data suggests that in that context, its performance is usually not as high as that of ultrasound. And so, most people would say its performance is somewhere in the range of maybe more like 20 to 40% sensitivity. So, I think the recommendation is to use ultrasound alone or ultrasound plus AFP with the more recent guidance tending to encourage the use of AFP with ultrasound. We have another question. So, what are our thoughts about the recent data that HCCs with mutations in beta-catenin are relatively resistant to immune checkpoint inhibitors? Is the evidence robust enough to justify checking for these mutations by biopsy or liquid biopsy? So, this refers to two studies, one published in Clinical Cancer Research last year from the guys at Memorial Sloan Kettering that show that essentially all the patients that had beta-catenin mutations in the tumor did not respond to immune therapies. There's another large study that demonstrates that if you have beta-catenin mutations across different tumor types, not only HCC, the tumors are excluded. So, they're not infiltrated by immune cells, which would correspond or would be kind of a surrogate that these patients may not respond to immune-based therapies. So, there is suggestive data, but we have some data still unpublished that suggests also that this may not be entirely or exclusively the case. So, it seems that patients with activations of beta-catenin may be immune excluded, but there's something else that in addition to the beta-catenin mutation may eventually render them primary resistant to immune-based therapies. So, I don't think the data is mature enough to recommend essentially not treating patients with beta-catenin mutations with immune-based therapies. We have a final question. To what extent would a composite of methylation mutations, CTC detection, increase sensitivity to early-stage HCC, or are the same tumors release components or some tumors are simply silent? So, what are your thoughts, John? What do you think? Yeah. So, I think it's probably always a good idea to look at a combination of biomarkers and avoid the trap of sort of settling into your own favorite approach. So, if you think of, for instance, the multi-target stool DNA test for colon cancer screening, that's an assay that looks at DNA mutation, DNA methylation, and protein. So, I think looking at a combination of marker classes is a good idea. I'm not aware of any data that combines all three of those approaches. Augusto, are you? Or Judong? Not in HCC. So, the closest thing is the cancer sick study published in Science from the people of Johns Hopkins that for the HCC patients that they'd included, so this is a large study for early detection of different tumor types. For the HCC patients, they included ctDNA plus AFP. So, I agree that there may be some room for protein-based or combination of different, let's say, signals to maximize sensitivity. You've got to be careful that you may also negatively impact specificity, but I agree with you that it makes sense to maybe combine these different approaches. So, with these, there are no additional questions. We're going to finish the webinar. I want to thank the panelists, Dr. Kiesel, Dr. Yang, Dr. Roberts, the ASLD for their support, and also the attendees. Thank you.

liquid biopsy

ctDNA

CTCs

hepatocellular carcinoma

early detection

treatment response

DNA methylation

biomarker

sensitivity

clinical practice