Scientist Stories: Sharon Plon, Germline Cancer Genomics

Good afternoon, and welcome to our lecture series today. My name is Charles Rotimi, and I am the scientific director for NHGRI at NIH. And I'm really, you know, glad today and happy to introduce our only named lecture series. At our institute, the Jeffrey Trent Lectures in Cancer Research. And I'm going to say, you know, a couple of sentences about Dr. Trent, and he was going to come on and, you know, say a little more about himself and his work. Dr. Trent is president and research director of the Translational Genomic Research Institute, TGen, in Phoenix, Arizona. You know, prior to forming TGen in 2002, Dr. Trent served for 10 years as the scientific director of the National Human Genome Research Institute, and under his guidance, our Division of Intramural Research became an internationally recognized research center in human genetics. So, Dr. Trent, thank you.

Well, thank you very much, Dr. Rotimi. What a wonderful opportunity to be able to share with you today at this seminar. I would have hoped that this would have been an opportunity to see everyone in 3D rather than the 2D we've become so accustomed to, but nevertheless, I'm particularly delighted to be part of Dr. Plon's lecture. As her work actually aligns perfectly with what nearly 20 years ago I gave Francis Collins, who was then the NIH director, as a reason why I was choosing to leave the mothership, to leave the NIH in order to start a fledgling organization like TGen. And that was recognizing that the power that the NIH could bring forward for a common good project like the Human Genome Project was one thing, But bringing genomic information into clinical practice, that is literally being on the forefront of bringing clinical genomics into the care stream, seemed to me to be something where industry, academia, and government were going to be involved in a cooperative fashion, and a nonprofit institute would be best positioned to help the patients sitting in front of us today. That's my departure to TGen, which is now part of the City of Hope. So let me just end by saying again that the reason that I believe Dr. Plon is the perfect individual to highlight her work today is, as hard as it is to believe for some on this, this video, 20 years ago we really were looking at one gene at a time. And for the better part of a decade, most clinical groups were satisfied with small panels of genes, all the while knowing that we needed to look at every gene and ultimately every base pair if we wanted to maximize information for our patients. So as you'll hear from Sharon, she has led the use of comprehensive genomic profiling for childhood cancers, and in that regard, she is both a world leader and a true kindred spirit. So again, thank you, Dr. Rotimi, and thank you, Sharon, for giving this lecture.

Thank you, Dr. Trent, and I'm going to say just a little more about Professor Sharon Plone. She's a board-certified medical geneticist and a longstanding career cancer genetics researcher. Her research has led to the discovery of new cancer susceptibility genes and the implementation of genomic testing in medicine. She holds the Dan L. Duncan Comprehensive Cancer Center Professorship at Baylor College of Medicine in the Department of Pediatrics, Hematology/Oncology, Molecular and Human Genetics, and Human Genome Sequence Center. She was the co-PI of the NHGRI/NCI-funded BASIC-3 clinical trial on the incorporation of exome sequencing into the care of newly diagnosed childhood cancer patients., a study that is expanding into diverse patient populations across Texas. She has served for almost a decade as one of the PIs of the Clinical Genome, the ClinGen resource, and chairs the ClinGen hereditary cancer efforts. She's currently co-chairing the germline reporting effort of the National Cancer Corps Pediatric MATCH Precision Oncology Trial. Dr. Plone is also working on a population-based study to understand the association between birth defects and cancer risk in children. I serve with Dr. Plone on the board of directors of the American Society of Human Genetics, and she also served on the NIH Human Genome Research Advisory Council. So the title of our talk today is Evolution of Germline Cancer Genomics: From Rare Disorders to Precision Oncology Trials. Thank you, and it's a wonderful pleasure to have you today.

Thank you very much. And as I share my screen, I just wanted to remind folks that if you have questions, please use the Q&A. I'm going to try to finish in enough time. To allow us to have some good questions at the end. So you've already heard all about me, so which I'm just incredibly honored to have this. And as Jeffrey— Dr. Trent said, he and I work in kindred spaces. I've had the pleasure of participating in the TGen Pediatric Cancer Genomics meeting before the pandemic and admire his work. I don't have any disclosures. I'm not going to talk much about rare disorders because I wanted to have enough time to talk about what's gone on in the last 10 years. But where were we in 2010? As you heard, I'm a medical geneticist. Prior to 2010, the way we figured out which cancer patients got genetic services were particularly in pediatric realm. A child had specific birth defects and a childhood cancer diagnosis, hemihypertrophy, things like that. Obviously, you're all aware of the work, hereditary breast ovarian cancer, Lynch syndrome, so individuals with personal and family history of cancer. There's an increasing number of cancers that we recognize that that cancer diagnosis alone is associated with a high likelihood of genetic predisposition. Retinoblastoma being the classic one, but medullary thyroid cancer and many others. Then unfortunately, if a person developed more than one malignancy, and all of those things would get you a referral to a cancer genetic service, you'd get a full genetics evaluation, you'd have pre and post-test counseling and testing. As was mentioned by Dr. Trent, we would often do genetic testing one gene at a time. But patients who didn't meet these criteria really didn't get any genetic services. So it was sort of an all or none, and that was something that always really worried me. What were we missing in the individuals who did not have that? I did just want to pull up one of Geoff's many papers in this area, and just— and this paper was published in Nature around 2010 and really highlighted sort of the revolution in research genomics. That was going on. And even back then, he and his colleagues Kevin Brown and others were already doing whole genome sequencing of probands, looking for new cancer susceptibility genes. And this is the classic paper that identified the MITF variant and familial melanoma. But at Baylor, what we really started to think about was the question of precision oncology and precision medicine. Precision oncology is defined as the tailoring of cancer prevention and treatment based on the characteristics of each patient. This was a slide my colleague Will Parsons put together many years ago. Of course, it's not a new concept. Oncologists have been doing that for decades. But the code words here, when you refer to prevention, typically you're talking about germline analysis, and for treatment, you're talking about tumor analysis. And what we started to talk a lot about at Baylor was the fact that to do precision oncology, you had to have clinical-grade genomic testing. You had to implement it. You had to put it into the medical record and report it back to physicians and patients in order to actually have precision oncology work. So one of the nice things about being a physician scientist is you have different parts of your careers, and one of the really sort of fabulous things I got to participate in was the early work starting in late 2010, early 2011 at Baylor to take the research exomes and translate them into the clinic. So Christine Ng and Richard Gibbs really led the effort. Ya-Ping Yang and Donna Musny did a huge amount of work. And a group of us, and we called ourselves, sorry, the management team, met every Friday for 2 hours for about 2 and a half years. To develop our clinical exome test. And we first applied it really clinically. We made it available to physicians to order, and we had the privilege of publishing our first 250 patients in the New England Journal. And around that time, as I'll mention in a minute, the Clinical Sequencing, or CSER, Consortium RFA came out, and that really changed much of my career trajectory. But of course, on the germline side, we went on by 2014 to publish our first 2,000 exomes. And then Jennifer Posey, which many of you know through the Greger program, actually when she was a fellow, worked with me and did the analysis on our first adult who had exome sequencing. So the availability of that test and our rapid translation to a tumor exome really in our work was really our— the onset of clinical cancer genomics. And we were able to ask the question, and many others as well, what happens when all cancer patients, children or adults, have access to genetic or genomic testing? So as I mentioned, NHGRI was at the forefront of this. They put out what eventually became known as CSER, the Clinical Sequencing Exploratory Research RFA. Eventually under the Genomic Medicine Division. And I did just want to do a shout out to Terri Manolio, who's head of that division, for being awarded the Presidential Rank Award yesterday. But we— well, Parsons and I, with the group I just showed you, said, you know, let's do this for pediatric cancer. Let's take that clinical exome test and offer it to every child, tumor and germline. And really explore what's going to happen. And we were one of the early CSER sites along with the— and there were two others actually doing tumor normal sequencing, University of Michigan doing children and adults with rare tumors, predominantly sarcomas, and the Dana-Farber doing adults with advanced malignancies. And so our project, as you heard, was called BASIC-3. Will Parsons was co-PI with myself, and Amy McGuire and her Center for Medical Ethics and Health Policy played a major role throughout. And we called it BASIC-3, really BASIC cubed, but that's because I was an MIT nerd, but everyone refers to it as BASIC-3. And what we proposed to do was essentially take all comers, all newly diagnosed solid tumor patients, not leukemia and lymphoma, but with both brain and non-CNS solid tumors, do the blood exome, do a blood sample for the germline exome test I just mentioned, We developed a tumor exome test, subtracting the germline and reporting that. So every patient had a CLIA-certified germline exome report, somatic mutation report. Those went into the electronic medical record, and I'm talking back in late 2011. They were shared with their pediatric oncologist, and our study genetic counselors with the oncologist shared them with families. So we returned results from the beginning, and then we followed the patients for 2 years. And our study objectives, of course, were number 1, to see could we really do this, um, at Texas Children's Cancer Center, and then evaluate the impact on both the physicians and the families. And I should just say the oncologists have been subjects in our trial and have been amazing research participants and stakeholders, as well as the families that have participated. In our studies. So the methods have all been published a number of years ago, but I'm going to give you sort of the final results of the study. This paper came out when we were about halfway through. So first of all, because there's a lot of emphasis now on diversity and race and ethnicity, Houston, of course, is considered the most diverse city in the country, and Texas Children's Hospital is one of the primary caretakers of children in the city. And so we were committed from the beginning to enroll all of our patients. And we provided everything in Spanish and English. And this paper that came out in 2014 for the first 100 patients, I've now updated to the whole study. And again, there were no statistical differences at all in who enrolled or declined to enroll. And so our study in the end was about almost 50-50. Self-described Hispanic and non-Hispanic. This is all self-described information. And by race, approximately 10% are African American. There's a few more percent African American in individuals with more than one race. And that very much represents what the state of Texas looks like. So this really was representative of our city and of our state. We also made what I think at the time was a fairly brave decision to return lots of types of results. So on the tumor side, we returned all somatic mutations. And on the germline side, we returned pathogenic and likely pathogenic in VUS related to the patient's cancer. And we included any cancer susceptibility gene or if they had another phenotype. We— this predated actually the ACMG secondary findings, but we had our own initial list, and then we adopted that list of other medically actionable findings. We had a small pharmacogenomic set of genes, and we actually returned all pathogenic variants in recessive OMIM genes. But I'm only going to talk about the results related to cancer. So again, I mentioned that we had a diverse patient population, and because this was started in 2011, we had a lot of variants of uncertain significance, but we were able to look whether that varied by race or ethnicity. And this was really some of the first data to show that we really had no difference in the number of VUS reports or median number of VUSs on the reports between Hispanics and non-Hispanics. But as has been reported by many others, we did have significantly more for individuals who self-described as African American. In terms of diagnoses, This was one of the early studies that showed that approximately 10%, almost exactly 10%, of the patients in the study had a molecular diagnosis. And by that, what I mean is they had a pathogenic or likely pathogenic variant in a dominant cancer susceptibility gene, or they had biallelic variants in autosomal recessive. I'm not including here patients who had a single recessive variant. So let me talk about the recessive first. My early career, I spent a long time studying rare recessive disorders like Rothman-Thomson syndrome. I'm disappointed to say that less than 1% of our patients— we only had one patient who had biallelic variants in a recessive gene. Um, and in fact, I'll show you data later about other studies, but I can just say now that all of them have reported the same thing, that if you look at solid tumor patients, much less than 1% of those individuals appear to have a recessive cancer cancer susceptibility disorder. It probably is higher for leukemia lymphoma patients because of diseases like Fanconi anemia. If you have one of those disorders like Rothman-Thompson syndrome, you have a very high risk of cancer, but they make up a very small proportion of the pediatric cancer population. The other surprising results were that even though we only had 26 dominant disorders, that made up 19 different genes. Only p53 and VHL had 3 patients each. All of the other genes only represented 1 or 2 individuals. And the other surprise, which as I'll mention has now been repeated by, at the same time and subsequently by many others, is that we found many patients who had a molecular diagnosis but a gene not previously associated with that child's tumor. And I highlight BRCA1 and BRCA2 and PALB2 here because that— I'll raise that again later. But you can see it was a wide variety of genes that we didn't think were necessarily related and we didn't know were related to the tumor. But SMARCA4 is a good example. This is a patient who had neuroblastoma, complete loss of heterozygosity. And since we published this case, there are at least 3 or 4 additional patients And a group of us— I mean, across many studies— and a group of us are writing this up to suggest that perhaps neuroblastoma is part of this. So we may be expanding the phenotype for some of these genes. Of course, as a geneticist, the question is always, could we have predicted who was positive? So age in adult genetics is the classic feature that's associated with cancer predisposition. That's not true in children. Children diagnosed at very young ages were as likely. There's absolutely no statistical significance here based on the age of diagnosis. We also, again, could compare Hispanic versus non-Hispanic because, again, we're almost 50/50. And again, there was no statistically significant difference. And there was actually no difference— sorry, that didn't come out right— between CNS and non-CNS tumor types. I don't have a slide here, but I can also say for every patient in the study, we said at entry, so prospectively, based on what we know about the patient, age of diagnosis, tumor type, anything in the electronic medical record, would we have recommended genetic testing? And then we went back and looked for these 28 individuals, and for 40% of them, we did not recommend testing for the gene in which we found the mutation. So that's about all I wanted to say about BASIC3, except for to say that, of course, we weren't alone in this. And these are 3 other early studies. The first one is the Other CSER study from the University of Michigan, also reported 10%, also reported that many of the patients had no family history. Again, wide variety of genes. The large St. Jude retrospective study that was published in the New England Journal 1,100 patients, of course, 8.5%. Again, wide variety of genes. Again, BRCA2 came out of that. Then a large study a couple of years later, again, retrospective, almost 1,000 patients, again, 7-8% germline findings. Again, they actually did a statistical analysis and showed that actually BRCA2 was overrepresented in their population. All of these studies and many others have commented that many of the germline-positive patients do not meet the current criteria for genetic testing. This, of course, applies to adults as well. I had the honor of being part of the TCGA germline analysis. And again, a paper now published a number of years ago, we showed that 8%, again, carried pathogenic variants in cancer susceptibility genes. And there have been many, many such studies in adults. You may note that some of those studies have higher numbers. Those often are studies that are including single recessive variants. And about 6% of cancer patients carried one of those. So that explains some of the differences in numbers across studies. So of course, everyone, especially my ClinGen hat, are interested in the issue of variant reanalysis. So we have followed these patients, and Sarah Skolan, our lead counselor, and I were reviewing these results for this talk. The largest number of variants that were reclassified are actually in that group of secondary or incidental findings. And we had 3 pathogenic variants that were downgraded. This is primarily in very early patients in the study. This predated the exact database, or NOMAD, of course, predated that as well. And it also actually predated the Richards et al., 2015 classification rules, which really tighten up how do you call— what evidence do you have to evaluate to call something pathogenic. So we've actually had 3 of those downgraded. I did just see, in fact, I was supposed to see, I didn't see the patient, one of those patients for a different reason yesterday. I noted that despite the reevaluation and the fact that we had informed everyone, she was still having echocardiograms. It highlights that once you have a label, a genetic label, that it can be hard to change practice. We did have one patient who had a negative WES, who through the work of my colleagues at Baylor actually was found to have intellectual disability and congenital anomalies actually was one of 3 patients where they described a new and published a new genetic disorder not related to their cancer. And we had one patient with a cancer diagnostic finding that was downgraded, and that was in DKC1, dyskeratosis congenita gene. That patient had completely normal telomere length, and in fact, that variant is now likely benign in ClinVar. And we had 1 upgrade. So it's just to say that among 300 patients, we had a fair number of changes. And I think this highlights a point that my colleague Heidi Rehm and I made in an editorial a couple of years ago, which particularly people that had genetic testing early 2010, 2012, '14 is the group where you really need to reevaluate some of those variants because of these subsequent developments. So just let me briefly mention, of course, that's one of the goals of ClinGen. I'm not going to talk about ClinGen today. This is the last year of the CSER Consortium, so I thought it was a good time to focus on that. But of course ClinGen is really focused on the relevance of genes and variants for precision medicine and research, and I hope that last slide kind of pointed out why that's so important. And we have a large number of ClinGen expert panels. I co-lead the hereditary cancer efforts with my colleague Doug Stewart. On many genes that are relevant to both adult and pediatric cancer, and we have a number of new ones planned as we've just entered 5 additional years of funding. But with regard to the CSER Consortium, there was a second round of funding and we applied again, but we expanded our team now across much of the state of Texas. So we called that study Texas Kids CanSeq, and we learned that it's good to have a logo for your study, so we now have a logo, and we were fortunate to to be selected for funding. We've increased the diversity even further, so now we have a diversity of clinical sites. It includes a small clinic on the Rio Grande border in McAllen, Texas, a large private hospital, Cook Children's in Fort Worth, and then other academic centers in Texas. And again, expanded to an even greater proportion of medically underserved families, as well as, of course, diversity with a focus on Hispanic individuals. We also expanded the entry criteria to include not just newly diagnosed patients, but children who had recurrent or relapsed solid tumors, as well as lymphomas and rare histiocytic disorders. And we've just completed enrollment of over 630 patients. All 6 sites did participate, and we overall have greater than 65% diversity based on the CSER criteria. We did things a little bit differently in Kids Can Seek, and I'll just show a little bit of data since we haven't finished reporting everyone out. First of all, obviously the field has changed. Jeff Trent mentioned that as well. Targeted cancer panels are a very important tool, and we are directly comparing for every patient the results of a targeted cancer panel that was designed for pediatric cancer patients, at Texas Children's Hospital with more extensive genomic testing by exome. And we were going to do more pilot work with genomes, but I have to say COVID sort of got in the way of that. On the tumor side, one of the other things we learned— sorry— from BASIC-3 is that pediatric oncologists really do follow Evidence-based medicine. And if the patient has a tumor that would— they know they can treat with current therapy, they will not generally diverge based on genomic findings. And so we really limited the tumor analysis to patients in this high-risk group. So people that have a very bad diagnosis or prognosis even at diagnosis, like a glioblastoma, or children that have relapsed or recurred. And the other reason the panel was so important for that reason, and I want to highlight Anshumoy Roy here, who is a molecular pathologist, started as a fellow in BASIC3 and now is our lead molecular pathologist for kids' cancer, is that his laboratory at Texas Children's turns the panel around in about 10 to 12 days so that they can use that tumor information very rapidly to make treatment decisions. And in that arm of the study, we are— the children in the high-risk arm, we are comparing not just to exome, because we also learned from BASIC-3 that you really need to look at the RNA of the tumor. And I should say this panel on the tumor side is a DNA/RNA panel. We're comparing it to exome RNA-Seq and array that's being run at the Genome Center Clinical Lab. Donna Musny is the manager of that lab with Richard Gibbs and others, of course.. And we will be able to compare directly the outcomes of these two types of tests. So I can't give you summary data, but I just wanted to give you a couple of stories. So one tumor story, this is a paper that just came out in Case Reports. And this is a patient who had an ependymoma, but was discovered through the RNA-DNA panel to actually have an NTRK fusion. Ependymomas can be quite serious.. And this patient presented with widely disseminated— this is the tumor, but these little marks on the X-ray point out the fact that it had already spread down the spinal cord. And quite surprisingly— that is OK. Quite surprisingly, the RNA analysis showed that this patient actually has an NTRK fusion. And NTRK fusions are one of the miracles of genomic medicine. They are highly— tumors with these are highly responsive positive to larotrectinib. It's FDA approved for that indication. Um, they're almost never seen in ependymoma. So our colleagues Will Parsons and Stephen Mack and others did research methylation profiling. And in fact, ependymomas are generally found in the little circle to the left or to the right. But in fact, the arrow points out that this patient, the little black dot, actually from a methylation profiling looks more like an astrocytoma. But like other patients, this patient has been highly responsive to larotrectinib. And you can see that there was great diminution— can't say that word today— of both the main tumor and the spinal cord lesions. And as far as I'm aware, the patient actually continues to do well. In terms of the germline side, very similar to what we saw in BASIC-3 are these unexpected results in patients who don't meet criteria. So this is a 15-year-old, previously healthy male with a very large oropharyngeal spindle cell rhabdomyosarcoma. Has been treated with chemotherapy and radiation. No family history, and so would not have merited— warranted a referral to cancer genetics or any genetic testing based on current guidelines. However, when Ang Shamoy reported out the tumor panel, There was a MYOD1 driver mutation which is seen in these tumors, but there were actually pathogenic variants in two different cancer susceptibility genes, both of which have been reported in ClinVar as germline, and both of which were actually pretty close to a 50% variant allele fraction, which is what we often see in cancer susceptibility genes. There was concern that one or the other of these might be germline in fact, And when the germline exome and the panel, they both agree with each other, came out, in fact, it was the DICER1 variant. So this is an example of a patient who would have had no testing, but because of the study, in fact, was found to have DICER1 syndrome. Actually, it was an inherited variant, and other members of the family are now having DICER1 screening guidelines as recommended by Doug Stewart and others. So the other thing that I really haven't had a chance to talk at all about is that there's a big team of folks that are really looking at the implementation part of all of this. And so I just want to shout out to Sarah Skolan, our lead genetic counselor. When we designed KidsConSeq, we realized there's no way with— we were aiming for 600 to 1,000 patients at the time— that we would have enough genetic counselor time to do disclosures for every patient. And so sort of based on our results, we— and we used to— and one of the things we did in BASIC-3, we continued but in a structured templated way, which is the oncologists are notified by email, and this is the treating oncologist, when the germline results are available in the electronic medical record. And we provide them a structured summary of what we think— we meaning myself and genetic counselors— what the germline results mean, and what the plan is for disclosure and whether there are any recommendations that we would recommend you do for the patient. And we're trying to model what some groups use as an e-consult in this finding. If the patient has a significant finding, they have an in-person or telemedicine visit by the study genetic counselor. And of course, because of COVID it was really good that we set it up this way because we were able to continue disclosures by telemedicine through Throughout the pandemic, we now are back to offering in-person, although most of the families do select telemedicine. That's if you had a diagnostic finding of any type, secondary findings, or even carrier status. Even someone found to be cystic fibrosis carrier gets an in-person visit. The families then— and those are in English or Spanish at the parent preference. Then they're provided the letter packet following the disclosure. If you don't have a significant finding, so we call those non-significant findings, and we do include any report with only VUSs in the non-significant findings, they get a letter packet only. It has a counseling letter. They have a copy of the test results. There's educational materials that have been developed through the stakeholder working group of the CSER Consortium. They're mailed to the family in English or Spanish. And we emphasize in multiple places in the letter that they have access to a genetic counselor for questions. They can email us. They can call us. Actually, very few families do. So what Sarah has been leading, and she presented this work last month at NSGC, and this is an initial evaluation of the first about 150 parent surveys, is every parent gets a post-disclosure survey. And there are a set of questions that have to do with actually the mode of return of results. So the first piece of data here is, would you have preferred to receive your child's genetic test results in a different way? And in fact, if they got to meet a genetic counselor, 90% said no. If they only got the letter packet, still almost 75% said no. But of course, this difference is statistically significant. Almost a quarter of them would have preferred a different way. Probably more important is we ask about their subjective understanding. We're not giving them a genetics test. We're just asking them how well did they think they understood the results, and again, you see there's a significant difference here. The ones who met with the counselor said very well or well or somewhat, but if you now look at the ones who got the letter, it's only 60% who said very well or well, In fact, you now start seeing even 12% that said a little or not at all. The conclusion from this part of the study so far, and of course we'll get more data as we just finished the disclosures, the majority of the parents are satisfied with all three modes. This did include a few in-person and mostly telemedicine and letter. However, parents receiving results by letter certainly had lower perceived understanding and more unanswered questions. That was one of the other questions on the study and had a stronger preference for an alternative mode. Interestingly, there were no significant differences in satisfaction or understanding by ethnicity and language, whether they preferred English or Spanish, and whether there was a VUS on the report or not. So our data suggests that disclosure by mail with sort of passive access to genetic counseling may not be adequate and may need some form of active support, even if that's just a phone call a week later. By the clinical genetics team. I should say, in contrast, I've just evaluated this evidence and submitted it to the American College for next spring, is that the structured email to the physicians does seem to be sufficient to really mediate their downstream physician recommendations of what they do with the patients. So I do just want to do a shout out. Part of the reason I think there's been no difference between English and Spanish preference is really the amazing work of Amanda Gutierrez, Alba Rocinos, and many others in our study. But Amanda, really starting with Basic Three when she was, I think, right out of college, said, you know, we've got all these disclosure visits that are recorded and you've used interpreters. Let me really look at what— how the interpreters interpreted exomes. And her first paper was really looking at the language that they used and actually made some recommendations. Subsequently, she led an analysis of really looking how medical interpreters may bridge sociocultural gaps. And now in the second phase of the CSER consortium working with Nong Olinburg and many others, there's now a whole team. This paper I think just came out with recommendations from CSER on how to create accessible Spanish language materials and sort of the lessons that have been learned from these studies. So fortunately in January of 2020, we had a team meeting. And this was all of us together. Of course, the data lives on. And so the BASIC3 and KidsCanSeq data are available for use by the larger genomics community. BASIC3 exomes, tumor germline, and any research transcriptomes are in dbGaP, and that data is actually being used right now in ClinGen for pediatric cancer hotspot analysis. We submitted variants to ClinVar, and we're doing even more of that with KidsCanSeq. The Kids CanSeq exomes are going into ANVIL. And we were selected as part of the Gabriella Miller Kids First Project to have whole genomes of the BASIC3 parents and children. And we're now actually part of a long-read pilot. And that data is all available in TBADAC. We, of course, ourselves have done more research. And just to sort of highlight a paper from a few years ago based on a single BASIC3 patient, We are able to identify, led by Ankshamoy Roy and Will Parsons, internal tandem duplications that are the most common driver mutation in this very rare sarcoma of the kidney. I mentioned our one recessive diagnosis that turned out— we collaborated with two other groups and described three patients with this new deficiency syndrome and a high risk of hepatocellular carcinoma. And a graduate student— oh, that disappeared. Yes, a graduate student in my lab is analyzing the whole genomes from the basic 3 trios and, for example, has identified a tandem duplication that he's now trying to interpret whether this in some way may have driven the ependymoma in that patient. So the second study I wanted to talk about, and I'll probably end with that, is, okay, so this was all put it in the medical record, provide the information, to the oncologist and let them decide how to use it, particularly in the patients who relapse. Would that make them eligible for a clinical trial or not? But with the data from BASIC-3 and from the St. Jude's data and many others, that data was really used to motivate that the Pediatric MATCH trial, which was modeled after the adult MATCH precision oncology trial, would in fact include germline reporting. So really, in this trial, we're answering the question of, can we implement tumor germline reporting in a national treatment trial? So Pediatric MATCH, which is co-sponsored by the National Cancer Institute and the Children's Oncology Group, is a single-stage Phase 2 trial of genomically directed therapies for children with refractory solid tumors. So these are tumors that have failed all current treatments. They get biopsied and sequencing at relapse. They identify actionable mutations. They're selected into a study agent arm, and then those individuals are assessed for a response. Uh, relies on the Children's Oncology Group, which has over 200 research sites throughout the United States. Unlike pediatric cancer, almost 90% of children diagnosed with cancer in the United States are treated And over 100, I think 120 sites from COG are participating in the MATCH trial. So for MATCH, there's tumor and blood sample sequencing for every patient using the FFP tumor samples at relapse. They're processed in the Biopathology Center. They're using a platform similar to adult MATCH, which is a targeted panel, oncomine DNA and RNA mutation fusion panel. It's got over 140 genes, and they really identify through their software analysis pipeline mutations of interest, and the sequencing is performed at those centers. For the blood analysis, the reporting is being done by two labs, Baylor Genetics and Children's Hospital LA. They do a lot of germline reporting, and this was not done for the Adult MATCH trial. There was no germline analysis in Adult MATCH. And this will obviously provide information on genetic susceptibility to cancer. In terms of that, we've actually been able to collect blood samples from more than 99% of the patients at entry. Um, the analysis focuses— the tumor analysis, as I mentioned, focuses on actionable changes, so mutation hotspots, loss-of-function alleles. MATCH does not report deletions or splice site variants unless it's a known hotspot. For the splice sites, and the matched germline reportings reflect what was found on the tumor report. So we identify the genes on the report that are likely to be cancer susceptibility genes. That's shown here. There's a variant on the tumor report. We then know immediately whether it's found in the blood or not. We've been doing this now for, I think, 3 and a half years. We meet every Tuesday afternoon, um,. And we get online and we review the study data through the Matchbox platform. So for the tumor, you can see the copy number for all the genes on the, on the panel, as well as if there are any genetic changes identified. The red and green little boxes here tell you if they match a treatment arm. And this is a patient, if you can't see it, that has two NF1 loss-of-function variants, and an MSH6. So all three of these could have been germline and related to cancer susceptibility. And in fact, one of the NF1 variants was— this patient with no neurofibromatosis— and you're essentially seeing the second hit in the tumor. Conversely, we have patients who have TP53 alleles. This again is a known variant that can be germline or tumor. The allele frequency is 0.976, so clearly there's loss of heterozygosity. But the concern was still whether this might represent Li-Fraumeni syndrome in a patient with a sarcoma. And under the blood result, it says no variant data. So this variant was not found. So I reported at AACR an interim analysis when we've completed 868 patients. Overall, 7%, essentially 6.8%, are positive for such a germline finding. It's an interesting step-down pattern. If you look at the 868 patients, about a quarter of them have a finding on the tumor report that could be suggestive of cancer susceptibility. TP53 or MSH6 or NF1. I just wanted to point out that the proportion of patients with a finding in their tumor that's found in the germline does vary by gene. And so in fact, what you can see here is surprisingly there are 22 patients— and this is when we were about two-thirds of the way done— who had NF1 findings, about a third of which actually have neurofibromatosis. Again, for RB1, it was about a quarter. P53 is clearly the big player in terms of variants in the tumor that could be obviously of concern, and there it's only about 15-16% that are actually in the germline. Interestingly, there are 20 patients with ALK missense variants that are associated with familial neuroblastoma. None of those have been in the germline so far, similarly for PTEN. But for BRCA mutations, and here I sort of cheated, I combined 4 genes, BRCA1, BRCA2, PALB2, and CHEK2, similar to what we saw in BASIC3 and have been seeing in KidsCanSee, Actually, if you see those in a pediatric tumor, almost all of them are germline. In fact, these represent patients with these hereditary breast cancer mutations. I just wanted to— I think I will finish here because I'd like to leave time for questions. If you think about it, what MATCH has sort of told us is that you can do paired tumor-normal testing for precision oncology trials. It's really the first national trial to perform prospective MATCH tumor blood analysis with the germline status returned to oncologists and patients. And I say that every, every patient in the trial gets a germline report. You don't only get a report If there's a finding, it obviously allows us to rapidly delineate germline status when you see it in the tumor. We've identified susceptibility syndromes in children and their family members that would be missed by current guidelines. And, you know, a point that my colleague Steve Joffe, who co-chairs the committee with me, has made many times is that we are providing useful information for the families in the trial, even if a treatment arm is not available. An interesting finding of the results being 7%, and I'll just highlight that again is an underestimate because we're not reporting deletions, we're not reporting splice sites, but it's still comparable to what's been seen in other, other studies. So it does suggest that pediatric cancer patients with recurrence or relapsed tumors have similar germline findings to studies of unselected pediatric cancer cohorts. So overall, bad tumors don't seem more likely to be germline. I will just say we do have some data that in specific tumor types that may not be the case. And here, because of the problem with my slides, I am just going to skip— first of all, sorry, I don't want to skip that. I want to acknowledge the huge number of people that have made the NCI COG Pediatric MATCH Study possible, in particular, Nita Seibel from NCI and Will Parsons. And again, the data I shared today is all data that's in our abstract from AACR. And I'm going to skip all this and just come to this last slide that ASHG this year, we looked specifically at a study we published last year at rhabdomyosarcoma. We then— Bailey, a graduate student in— Bailey Martin Giaccoloni, a graduate student in Philip Lupo's lab, one of my collaborators did a really nice analysis that looked at overall survival and event-free survival. This study was also done with the Children's Oncology Group, so we do have really good treatment data. And in this case, what we found in rhabdomyosarcoma is, again, it's about 7% of the patients, all comers, have a germline variant. If you limit it to embryonal rhabdomyosarcoma, it's about 10%.. And the group that have germline findings actually do significantly worse in terms of overall survival or event-free survival. And this is important because embryonal rhabdomyosarcoma is actually thought to be the better-acting rhabdomyosarcoma. And so the germline status in this setting may help us to identify patients at diagnosis who actually have a, uh, poorer survival, might eventually down the road need alternative treatment. And these are patients who don't have the PACS3 or 7-FOXO1 fusion that drive unfavorable outcomes in the alveolar form. So with that, I hope I've sort of convinced you that 10 years after the first tumor exome, we have the early realization of precision oncology for childhood cancer, and much of the same work's being done in adult cancer. Multiple studies suggest that 8 to 10% of pediatric and adult cancer patients have a monogenic cancer predisposition syndrome. Current testing guidelines miss a substantial fraction of these findings. And really, I hope I've sort of convinced you that one path forward may be offering germline testing to all pediatric cancer patients. I didn't have time to talk about BRCA2, but our prior assessment of adult versus pediatric cancer predisposition genes clearly needs to be reevaluated. There's much more overlap in that than I think was expected. I hope I've convinced you that we demonstrated the ability, much with the support of the CSER Consortium to conduct— and with the MATCH trial— to conduct precision oncology treatment trials that efficiently provide tumor germline reporting. And this may serve as a model for how to do rapid reflex testing for tumor early test results. And my last slide was trying to convey that there's increasing evidence that germline data may provide important information for outcome for a cancer patient or eligibility for targeted trials. And with that, I will stop sharing and turn it back over to my NHGRI colleagues.

Very impressive presentation of clinical genomics and precision oncology in action. So I'm Paul Liu, the Deputy Scientific Director of NHGRI, so I'm happy to moderate the question and answer session for this lecture. So the first question comes from Dr. Eric Green, the director of NHGRI, and he says, Sharon, thanks so much for giving this year's Trent Lecture and for giving such a fabulous presentation. You're truly a role model as a physician scientist. Your work on rare diseases and cancer genomics clearly illustrates the changing landscape in terms of the use of genomics in medicine. And he has two questions. First one is, what are your thoughts about how medical schools are training future physicians in terms of knowledge and awareness of how genomics will influence the practice of medicine in the coming years?

Yeah, it's a great question, and I just finished lecturing actually on cancer genetics, but of course it was this year by Zoom. I would say that I think medical schools are trying. It's a very tough question and there's still a lot of focus on classic cancer genetics, retinoblastoma and things like that on board exams, and so that translates into it. I will say Baylor has had a unique pathway, and a shout-out to Lori Pataki, Shweta Dhar, and others. We have a We have essentially an interest group where medical students who are interested in genetics can do a number of additional genetic activities and get a certificate for that. I think that medical students are very interested in precision medicine, but they're not necessarily seeing it modeled when they do their rotations, and I think that's where we're likely to have the biggest input. Whether that's doing sample cases or that kind of thing.

I think they really need to see it influencing patient care to sort of drink the Kool-Aid.

Do you think the faculty need to be strengthened in this area too, I guess? Well, I think that that's— I mean, you need the faculty to be strengthened and applying it in order for the students to buy into it, but I do think Perhaps all the emphasis on COVID and messenger RNA and vaccines will, in and of itself, create some additional interest in this area.

Okay. The second question from Eric is, are you keeping pace with the opportunities in genomic medicine that you described in your talk?

Well, I haven't applied for the genomic medicine RFA, if that's what he means, but I've encouraged Sarah Scollin to do so. Hopefully I'm keeping pace.

I think I'll leave it at that. Here's one from an anonymous attendee. When calling the biallelic recessive cases, did you include trans compound heterozygous?

What would occur? This is one of the advantages of doing exomes with kids, which is that we got samples from parents. And so any pathogenic variants in recessive genes, the lab— if there were two, the lab would test the parents so we could set phase. Um, we have had a couple of patients who have VUSs in cancer susceptibility genes in both genes, and you can maybe argue about that. But as I said, there have now been a number of studies that have tried to look. There are very few— I mean, they exist certainly at St. Jude, presented someone with ataxia telangiectasia.

There's, you know, it's just they don't make up a large proportion of childhood cancer patients. Yeah, seems like very few. A question from Les Biesecker: Earlier in the talk, you showed the finding of surprising germline variants in genes not predicted by the patient's tumor. Does that have implications for secondary findings in patients with pediatric cancer?

Yeah, and so the slides I skipped over, the best example of that is BRCA2. So we only reported one in BASIC-3, but in our rhabdomyosarcoma study, we reported 6 patients out of 600, so 1%, and it was clearly one of the only non-rhabdomyosarcoma genes that was significantly overrepresented in controls. My graduate student Adam Weinstein has really reviewed the world's literature on this, or at least 4 other studies that have showed that BRCA2 is overrepresented in pediatric cancer, and specifically BRCA2, not BRCA1. And a couple of studies have showed PALB2, which is the partner of BRCA2, is overrepresented. So I think that's an example where we may think of it as a secondary finding, where in fact, it may be playing some role. And we're certainly interested in doing—

Adam is He's just working his thesis on that question. So based on this, probably we should look at more kind of broadly rather than limiting to only genes relevant to certain cancer?

Yeah, and so for example, the original pediatric cancer panel at Texas Children's did not include the mismatch repair genes, and then it became clear not only is there the rare recessive condition, which we have diagnosed one of those patients in Kids Can Teach, but in fact, Michael Walsh first reported, we have a paper under review, that about 0.5% to 1% of pediatric cancer patients appear to have Lynch syndrome. So that's another example where we added the— I mean, not me, but Dr. Roy's team added the mismatch repair genes back to the pediatric cancer panel. So I do think limiting the panels can—

will prevent us from learning about some of these relationships. Okay, next one, hopefully a short question from Yiyi Liu. For the MATCH trial, do you report variants with uncertain significance?

Generally, we don't, and that's because the MATCH— the MATCHBOX pipeline does not call them out on the tumor report. There's a few cases where something on the tumor report is actually a VUS in germline status, and those are the only ones we report.

Reported. Okay. Next one is from Leah Cunningham. I hope I read this correctly. Wonderful talk. I got the sense that your statement, "bad tumors" quote unquote, don't seem more likely to be germline, pertain primarily to solid tumors and not liquid tumors.

Is that your statement? Yeah. I should have had a caveat from the beginning that I did not show any data related to leukemia today. There definitely are examples in leukemia, and I tried to show the example in rhabdomyosarcoma where having germline status is a poor prognostic feature. I think we went into MATCH thinking that, okay, everyone in the study had to have failed treatment, that we might have a much higher rate of germline findings than, uh, that have been reported in just unselected patients. And there we don't see it, but there are definitely individual tumor types, and leukemia is a good example where germline status can be a poor prognostic factor.

Right. And also your embryonic, yeah, RNAs, of course, is an example. So we're running out of time, but there are still some very good questions. And Dr. Alter, Blanket Alter, what is the cancer significance of a single mutation, for example, in BRCA2 in a child with cancer versus two mutations of BRCA2 equals Fanconi anemia?

With a specific cancer spectrum? Well, hello, Dr. Alter. And I avoided saying anything about single Fanconi alleles today, but the— I think it's sort of what I said in answer to the other question. We initially thought, in fact, I think we even wrote in the BASIC3 paper that this was probably a coincidence. I would say there's enough data now that suggests single pathogenic BRCA2 variants are not a coincidence. But they have been reported in medulloblastoma, rhabdomyosarcoma, lymphoma, and osteosarcoma. So that is a different pattern than you see in the FAD1 patients.

But happy to talk to Dr. Alter about that further. That's good. One last question, then we'll have to finish. Although you mentioned that there was not a correlation of age with germline mutations in pediatric cancers as a group, has an age correlation appeared when considering individual pediatric cancer categories?

Well, I mean, of course there's some classic ones like retinoblastoma where it's well reported, but there are other examples. Gail Tomlinson's reported that hepatoblastoma in patients with APC are not necessarily younger, the patients with germline finding. So I think there are some other individual tumors where age at diagnosis has not been a good predictor in pediatric cancer.

Very nice. Thank you very much for, you know, all the answers to the questions.

Well, thank you for the great, great questions and the great honor.

Thank you very much.

Thank you. Bye-bye.