Intro: Welcome to Precision Pathology Podcast, featuring interviews with authors and global thought leaders who are at the forefront of technological advances in precision diagnostics and therapeutics. Your host is Dr. George Netto, the Simon Flexner Professor and Chair of Pathology and the Laboratory Medicine at the University of Pennsylvania's Perelman School of Medicine in Philadelphia. Here is Dr. Netto.

George Netto, MD (Host): Welcome to a new episode of P Cubed. My guest today is Dr. Felix Sahm. He's a Professor of Pathology, actually neuropathology to be more specific, and a Vice Chair of the Department of Neuropathology at Heidelberg University Hospital. And he's also the Director of Molecular Neuropathology in that hospital. And as I mentioned, he's a professor in the university, which, I just learned, is the oldest university in Germany. So, we were kind of bragging and comparing notes about UPenn versus Heidelberg.

It's really a pleasure to have Felix. I'll say just a couple of words about his impressive background. Felix has over 300, somewhere I saw 400, but at least 300 publications. And these are all in prominent journals, very high impact journals, including the topic of our conversation today from Nature Medicine. More importantly, Felix has been an important player in the field with patents on methylation profiling and on the technology and neuropath. And he is a member of the WHO Working Group in neuropath that helped us with classification. He has involvement in ESMO, writing ESMO guideline. So really, an exemplary precision neuropathologist that's really putting our name on the map.

Today, as I mentioned, we're going to discuss a study that he led in a multi-center validation of a platform around doing intraoperative rapid molecular profiling of brain tumor, believe it or not. The timeframe described in the paper is 30 minutes. We have a hard time getting regular frozen sections in 30 minutes. So here, potentially, we can classify the tumor, and I'll let him talk about that. So, really, really, very advanced novel technology. Thank you, Felix, for accepting the invite. And I'm really excited to talk about the paper.

Felix Sahm, MD, PhD, MBA: Yeah. Thank you. It's my pleasure to be here with you today, George, and to talk about our research, or should I say actually about our practice because we have meanwhile transferred that already pretty close to application for patients. But I will talk about that later.

So just for the beginning, you said that it is Interoperative. Yes, that is true. But in the beginning, that was not our aim. So, we thought of something else before. We first thought, "How can we put all of the different markers that the new WHO classification now recommends, or even mandates since 2021 for brain tumors? How can we put all of that into one assay?" Because we have methylation now, especially in brain tumor diagnostics, also for sarcomas in some instances. We even have some tumor types in the CNS classification where methylation is an essential criterion. So, it's really needed in daily practice. But of course, we also have a lot of single nucleotide variants that are important for diagnostics, for treatment in physician oncology, obviously; gene fusions, and so on and so on. And all of that currently is done in separate tests. And therefore, there's an immense workload for a neuropathology laboratory, obviously, also for general pathology.

And that somehow elicited a debate about the applicability actually of the World Health Organization classification in the entire world, because many countries, many communities, colleagues told us, "Well, that is not applicable actually for centers outside Europe and North America." And that was actually our first intent to put all of this into a setting where you do not need like a million-dollar device for sequencing and half a million-dollar device for the next test and so on. So, that is why you put all of that into one assay. And then, later, we figured out it can also be pretty fast.

Host: Excellent. So, you touched on a very important point, the mandate of WHO. If you keep the classification, keep evolving into molecular and brain CNS, led with the methylation and keep adding essential criteria, it's making it very hard for a wide swath of the globe to use that classification.

So, we'll see, hopefully, this-- even if it's sometimes a hundred thousand dollars, it's still very expensive for certain communities. Even in the U.S., not every lab is a reference lab. So, let's dive in. We talked about the background, the need. So, how did you approach it? What is novel about this?

Felix Sahm, MD, PhD, MBA: Yeah. So in the first part of the paper, we are leveraging this long-grid sequencing technology, specifically nanopore sequencing. And the advantage of the nanopore sequencing is that it works totally different from all the other traditional sequencing that we are using. So, different from Sanger sequencing, but also different from the short-grid sequencing that most of us use in daily practice, so NGS. It does not sequence the DNA by replicating it, like we do it on a flow cell or like we did it with Sanger sequencing before, but it just takes the native DNA and pulls it through a pore; therefore, the name nanopore. So, it has a thousand of pores on the flow cell. And there is an electrical current, and that current is changed, depending on what base it's pulled through. So, it identifies what base is actually in the pore. And one base by a time, it identifies the sequence of the DNA fragment in every given pore. And it cannot even identify or not only identify four bases. It can also identify the methylation of the bases. And there, we have everything, right? So, we get a sequence and the methylation status in one assay.

Host: And to rephrase it for our audience, you're not having to amplify, you're not having to go all these additional steps, and fish certain segments of DNA out. And so, you're dealing with the native DNA and putting it through pores with that voltage variation. And you're saying which base this is and is there methylation or not. Sounds space age.

Felix Sahm, MD, PhD, MBA: Exactly. And we still do an enrichment. We do not do an enrichment before, we do the enrichment in real-time. And I just briefly explained that. So normally, for our NGS applications, we would do an amplicon approach or an hybrid capture approach to get the areas of interest. But here, we actually use a specific twist on the nanopore device. Doing the sequencing of a DNA fragment, we already align the sequence of the fragment that is in the pore before it is finished. So, we know what area of the DNA or of the genome actually is in the pore. And we have predefined a set of genes that are relevant for brain tumor diagnostics. IDH, BRAF, you name it. If the DNA fragment belongs to these interesting areas, the pore keeps on sequencing it. But if it turns out to be a fragment outside of the areas of interest for brain tumors, the voltage is changed and the DNA fragment is ejected. And the pore pulls another DNA fragment out of this entire lysate and, again, checks if there is interesting or not. And that means it is enriching in real-time without any library preparation before.

Host: So, it's focusing on really what's important for the tumor at hand, because you have those set of tumors. So, I guess that's what's called adaptive sampling-based sequencing. So, it's adapting. "Here's an area. This is of interest for brain tumors. Let me focus on that sequence, more native fragments of DNA." And then, "This is not, let me skip it."

Felix Sahm, MD, PhD, MBA: That's exactly the point. Yes. And you could also dynamically change it. So if there is a publication tomorrow that tells you, "Oh, there's another gene of interest in brain tumors," which does not happen so frequently anymore today. But if that would be the case, you could just change the table that is behind the program, and the algorithm would add another gene.

Host: So basically, the advantage of doing something like that is the rapidity, right? Because you can just focus on these spots, so you can do it in a faster way.

Felix Sahm, MD, PhD, MBA: Right. right. right. So, you save the entire time you would have for the library preparation. You just do a very small step that just makes the DNA applicable on the flow cell. But otherwise, the entire preparation other than DNA extraction, of course, the entire other preparation is just not needed anymore.

Host: Excellent. So potentially, you can do the profiling in 30 minutes you're mentioning, which is the span for intraoperative. But I realize your workflow for this study is two different phases. And the 30 minutes, not like you get everything. You're going to still the next day or next 24 hours get the comprehensive. So, can you explain to our audience what you get in the 30 minutes and how that guides your therapeutics or not?

Felix Sahm, MD, PhD, MBA: Of course. So, that's a good point. We actually have a two-step approach. It is one sequencing procedure, but we have a time point that is comparatively early, that can be 30 minutes, can be a little bit quicker, can also be a little bit longer where we get a methylation profile of about 90 different classes or a smaller set of methylation class that can be identified. And we got the copy number alterations that actually goes pretty, pretty quick sometimes in a few minutes. And it's especially helpful for like 1p/19q in oligodendroglioma, for 7/10 in glioblastoma and so on. So, this a rough methylation profile and copy number we get pretty quick. And in most cases, that is of course already highly informative. So if you have a diffuse glioma in the frozen section and you see it has 7/10, that goes to glioblastoma. If you see it has  1p/19q that would rather go to  oligodendroglioma and so on. So, that is very useful.

And even for meningioma, you have talked in the other podcast about meningioma, about changes like  1p deletion And so on, or CDKN2A so that you also could possibly already see early, and that makes a difference, of course, for the surgeon to know that. And if you then just keep on sequencing maybe two hours, four hours or so, then we also reach the read depth that we need in diffuse glioma to call SNVs. That is not possible yet in 30 minutes because the read depth is just not sufficient, especially in this diffuse tumors. But then, we can see if there is IDH mutant, if there is third mutation, even that in the intergenic region of the promoter, we can even see the third promoter mutations. And because of the long read technology, we sometimes even capture structural variants, that with conventional sequencing, are harder to detect because the reads are shorter. And if there are structural variations, then it is harder to align them. And we've seen that in the manuscript data also, that can even be superior to the conventional sequencing.

Host: Very impressive. So, that second phase, so the first step is enough to inform you, is this an oligo, is this GBM, and like you mentioned, 91 class-- I'm learning here, you guys in Heidelberg. So, you developed a 184 class is the final target so that you can get to later in few hours. And those long deletion-- because there's long reads, you're looking at structural variants, you're getting fusions, you're getting SMVs, like the mutations you mentioned-- wow, impressive-- and all that is possible as a second phase.

By the way, I just wanted to make sure for the audience that the study encompassed 301 samples, right? Including 18 samples that were sequenced intraoperatively.

Felix Sahm, MD, PhD, MBA: Yes.

Host: And these are, in my understanding, from Germany, of course, Heidelberg and others, and from Norway, UK and one in the U.S. Is it Chicago? Am I correct, in the U.S.?

Felix Sahm, MD, PhD, MBA: We had cases from different sites in the U.S. For example, from Cincinnati, we had cases from Oslo. And very instrumental in our project was a group that we collaborate with in Nottingham. They also had actual intraoperative samples. And from the other partners, more or less across the world, we received data that was not generated only on nanopore. We also received data that were generated on other methylation platforms.

Host: We'll talk about that in a second. So, this is awesome. So, this is what you just described, that's what you called it, the Rapid CNS platform that gives you that 30 minutes. But reminded me now that let's talk about the second platform in the paper or the second tool that you developed. You called it MNP-Flex, methylation neuropathology. So, that was our initial MNP, the Heidelberg system. And this is the MNP-Flex. So, the flex element is what you're trying to describe, that it's amenable to data generated by any platforms. Correct?

Felix Sahm, MD, PhD, MBA: Exactly, exactly. So, the initial Heidelberg classifier was trained and basically completely developed on arrays, methylation arrays, and we have like more than a hundred thousand data points meanwhile from methylation arrays. But while some neuropathology centers or also other centers have this methylation array system and the scanner and so on, there are of course many more institutions that have a sequencer, but not the array scanner system. Some institutions that have already bisulfite panels or that have any other technologies for methylation analysis. And we thought to make that more available and also more accessible for colleagues that may have sequencing devices, but nothing else so far. We developed a system that can take in all kinds of methylation data, as you say. So  bisulfite-converted DNA, also DNA from the array, of course, nanopore sequencing data, so whatever gives you any methylation information. And this we then called a flexible version of the MNP classifier; and, therefore, in short, MNP-Flex.

Host: So basically, if somebody just puts that data into-- so I imagine it is web-based and similar to the others, where you can dump that data that you have on the methylation and it will help you classify.

Felix Sahm, MD, PhD, MBA: Exactly right. So, there is an online version available. And as always, in such instances, it's of course important to say that this version currently is for research use. Of course, we are validating that we are developing that to a certified product, but that is now pretty fresh research product, obviously from the publication. But we see already quite some interest from colleagues And we also for our collaboration partners provide that so that they can use it locally. And we have already seen first publications subsequent to ours that come up where colleagues independently of us just use this tool and have pretty nice results. Also, for example, on liquid samples with this technology.

Host: Liquid CNS? Fluid?

Felix Sahm, MD, PhD, MBA: Yes.

Host: Wow. So, you plug it in and you get their profile.

Felix Sahm, MD, PhD, MBA: So, it's still challenging. And I have also not seen much data on plasmas, liquid biopsy from CNS tumors. But there are some studies that actually show that there's just less DNA shedded into the blood compared to the CSF. But from the CSF, other colleagues as mentioned, but also in our studies we see that it is is very promising.

Host: Wonderful. Wow. This is very impressive It sounds like it's going to be a very useful tool and simplify things. Just without mentioning numbers, is the nanopore technology, is that more expensive or less expensive? Let's say a place doesn't have that technology, how does it compare, you know, to the existing existing technology?

Felix Sahm, MD, PhD, MBA: Yeah, that's a good question. And that is not so easily answered, because there are two different components that is the initial expense for the devices. And then, of course, the running cost for individual samples. The initial capital expense for the device is in a, well, low percentage of conventional sequences. However, the region cost on a running basis, that is actually not much different from what you would have otherwise. Of course, it depends on what you really compare, what sequencing depth you compare, what kind of lability of CNVs you compare, and so on. But the reaction costs are not like a dimension different from what we would've otherwise.

Host: Excellent. So potentially, and because we started this conversation trying to democratize these amazing discoveries that you and other teams are making. So, the idea is we want the entire globe to be able to capitalize on this precision personalized diagnostics that will impact outcome. So, it sounds like at least you can do a reference center without inordinate amount of investment, at least that first initial. Even though it doesn't have to be fully granular, but it guides the management of the patients in countries where the full technology is not available. Is that how you see it? Do you see the future?

Felix Sahm, MD, PhD, MBA: Exactly. Exactly. And we see that already developing. So, we have founded a consortium, which we call the MNP International Outreach Consortium, and we have partner institutions from literally all around the world, so from Chile, from Brazil, from Argentina, but also from Indonesia, from India, and so on. We have already, for three years now, every year, a workshop where people from all these countries come to Heidelberg, where we discuss with them, where we actually can make these technologies, but also other physician oncology technologies more available, more accessible. And we now have in several of these countries, at least one national hub where this technology and this kind of novelty of brain tumor diagnostics is now available for the humanities.

Host: That is wonderful. And of course, being online for other countries who have the technology, but they can plug in the AI part or the classifier part and take it the next step. How wonderful, really, Felix, and your team. I know Heidelberg led the word in this, and now a lot of us including here, and I know several places in the U.S. are trying to develop this. And of course, classifiers need always validations and more cases you do. But really, congratulations on a huge impact for these difficult tumors. And to help these patients is something you should be very proud of. Anything else you want to mention about? And you touched upon Matija Snuderl, right, last meningioma podcast we have. So, it's not totally fortuitous that we had back to back CNS discussion, because there's a lot of advances in CNS and it's a welcome advance, because finally we can push the envelope in treating these patients. Anything else you want to tell our audience about what's down the road in your shop or other places?

Felix Sahm, MD, PhD, MBA: Yeah. I think what is currently a hot topic, certainly beyond neuropathology, everywhere in our field is that we may even get quicker and also cheaper in diagnostics by applying AI on histology images. So, we have currently already a paper submitted in review where we have a high kind of variability of over a hundred brain tumor types that are identified just from an H&E section. And we are working on a similar project also for intraoperative application. So, just imagine you would have a frozen section and you not only get the information it is epidymoma or oligodendroglioma, but you get the information like it is a ZFTA fusion positive fusion positive epi just from the H&E. I think these are developments not only obviously, not only at our department, but all across our discipline that will be maybe more transformative even than what we have talked about.

Host: Absolutely true. That's the hope. And really, these are the topics we want to continue to talk about. And maybe we'll bring you back in the future once you publish these other papers. But to have these digital biomarkers that we can do, that will be ultimate democratization, right? Because even in middle to low resource countries, you can just snap an iPhone image of the H&E and compare through an AI, at least get some preliminary. That will be awesome. Thank you again, and it's really been a pleasure having you here.

Felix Sahm, MD, PhD, MBA: My pleasure

Host: And this happened to be our fifth episode this calendar year. And the holidays are coming. So, I want to wish you and our audience happy holidays. We have a lot to look forward to in the new year. More great speakers like you to come and visit us. And, hopefully, we build upon this first year success. Thank you very much.

Outro: We would like to extend our sincere gratitude to today's guests for their valuable contributions and thoughtful perspectives. We also wish to thank our dedicated production team and the technical team at DoctorPodcasting. Special thanks to Jennifer Beppler for her dedication and skill in coordinating our episodes.

Please note the opinions expressed by our speakers are their own and do not necessarily represent the Perelman School of Medicine or the University of Pennsylvania. Thank you for listening, and we look forward to welcoming you back for future episodes.