Episode 3.1 - Single‐cell DNA and surface protein

Charles De Bock (: 00:16

Hello and welcome to the HemaSphere podcast. My name is Charles de Bock and I'm your host. When it comes to improving survival for cancer patients, acute lymphoblastic leukemia stands out as the success story from essentially a universally fatal disease in the 1950s to now where we see over 90 % of children surviving thanks to risk adapted therapies and treatment intensification. But leukemia still remains one of the most common childhood cancers numerically and is still a leading cause of death.

And so today I am joined by Professor Jan Cools and Dr. Margo Aertgeerts from the Center of Cancer Biology, the V.I.B. Leuven, and Dr. Heidi Segers from the Department of Oncology, KU Leuven and the University Hospital, UZ Leuven. Welcome to you all, because today we are going to discuss your latest findings published in HemaSphere with the title, "Single-cell DNA and surface protein characterization

of high-hyperdiploid acute lymphoblastic leukemia at diagnosis and during treatment". Right, questions. So I have many because it's a very interesting paper. As I mentioned, the survival rates for children with acute lymphoblastic leukemia now sit above 90%. And of all the subtypes of B-ALL, high-hyperdiploid is the subtype that really has a great prognosis. So

Why did you focus on this patient subtype and what are some of the clinical challenges that remain for these children?

Heidi Segers (: 01:44

ALL is the most common type of cancer in children, accounting for approximately 25 % of all childhood malignancies, and 85 % of pediatric ALL cases are of precursor B-cell origin, or B-ALL. And treatment typically involves intensive multi-agent chemotherapy for about two years. In our current risk-based treatment protocols, children with a higher risk of relapse receive more intensive treatment.

to improve outcome, while those with a lower risk of relapse are given less intensive treatment to minimize toxicity. And the risk certification is currently based on clinical characteristics, biological features of the leukemia cell, and early treatment response, which is assessed using minimal residual disease. Thanks to international collaboration, improved supportive care,

and more precise risk stratification, the five-year overall survival rates for children reaches 90%. Despite these high survival rates, incurable relapses and treatment-related toxicities remain significant challenges. Around 10 to 20 % of patients experience relapse, often with a poor prognosis, and chemotherapy-induced acute and long-term toxicities affect the majority of patients.

Therefore, more refined risk stratification and personalized therapies, including novel agents such as targeted therapies and immunotherapy, are essential to further improve treatment efficacy and reduce toxicity. High-hyperdiploid ALL is indeed the most common subtype of pediatric B-ALL, representing approximately 25 % to 30 % of all pediatric B-ALL cases.

and it's generally associated with a favorable prognosis. However, due to its high prevalence, it's also responsible for the largest absolute number of relapses. One of our previous studies using single cell sequencing across multiple B-ALL subtypes identified most heterogeneity in small mutations within the high-hyperdiploid subtype.

And therefore, we want to investigate this high-hyperdiploid group in more detail using single-cell DNA sequencing. Our aim is to determine whether a subgroup within high-hyperdiploid B-ALL cases can be identified that carries a higher risk of relapse. This may help refine future treatment certification, guide treatment decisions, and potentially improve outcome.

Charles De Bock (: 04:18

Exactly so.

when you're talking about the toxicity, think there is an underestimation that when we talk about overall survival and we see 90%, we see that as being the ultimate goal. But obviously that comes with so many toxicities and I've heard pediatric clinicians in the hospital themselves, often they're walking the wards and treating the toxicity rather than the leukemia per se. So it's really great that you're using single cell analysis and perhaps I can focus a bit on that because you sequenced 13 patients in your current study.

and this was with the diagnostic samples. And when you looked at the clonal composition and the co-occurring and mutually exclusive mutations, so what were some of the surprising findings that you didn't know before and you found out with this particular study at the single cell level?

Margo (: 05:04

Yeah, so I think one of the most interesting things we found is that we could actually find RAS mutations, so either in the KRAS or NRAS gene is what we mostly saw, and it was in each of the patients present, sometimes in a very small subclone only, whereas think bulk sequencing or previous larger bulk sequencing studies have identified these RAS mutations only in about 50 % of the...

high-hyperdiploid cases. So the fact that we found them in all of them was rather surprising. We did indeed sometimes see a very small subclone, which can indeed not be picked up on with larger bulk sequencing. So that might explain why. And then actually we saw even in six out of 13 cases, there were multiple RAS mutations. So either two different KRAS mutations in a different subclone or a KRAS and an NRAS mutation or

or even more. So that was quite interesting to see. And then if we also looked at co-occurring or mutations that were not co-occurring, we could see that flitree mutations never seem to end up in the same subclone as a RAS mutation. So they seem mutually exclusive. And this, think, was also already

ation with a glutamic acid at: 1099

maybe working together more and we even found them I think in about 30 % of our cases whereas they were described only in 3 to 6 % of high-hyperdiploid B-ALL so we also thought this was surprising and interesting to see.

Charles De Bock (: 07:09

Yeah, because I guess when we think about FLT3 mutations, you often associate with AML in particular. And so it's nice to see that it's also here, which gives some, and going back to Heidi's comment about personalized medicine, that this obviously opens an avenue for personalized therapies, at least targeting the FLT3 mutations. So RAS mutations are nearly in every single patient, which...

you know, when I think about RAS mutations, they are some of the most potent oncogenes in cancer and very difficult to treat. So when high hyperdiploidy is so favorable in terms of prognostic outcomes, how do we reconcile this finding where RAS mutations are so prevalent in every patient, even if it's a subclone with their...

Great outcomes. Are we missing something there? Or maybe perhaps you can comment on that.

Jan Cools (: 08:03

Yeah, I think we don't understand yet what the RAS mutations are really doing in this high-hyperdiploid subgroup and why they are so prevalent. There are some other studies who have shown that in cells with abnormal chromosome numbers that RAS pathway activation seems to be required to stabilize these abnormal chromosomes. So that may be the reason why we see it also in this high-hyperdiploid cases, because they have a higher number of abnormal extra chromosomes.

The real meaning and especially why it is in a subclone is also puzzling. Sometimes we see it in a very minor subclone and it does not seem to become the major subclone. If a RAS mutations are so powerful, then you would expect that they would be the major subclone at the time of diagnosis and that's often not the case also. So somehow they seem to be needed in a way, but not maybe in the way that we think of it in a classical way and as a really strong oncogene. So maybe there's another reason here.

And it's true, I think in some relapse studies in ALL, RAS mutations have popped up, have some acquired mutations at relapse. But I think also here in high-hyperdiploid cases, in previous bulk sequencing studies, it has not been identified as a major oncogene that is more frequent in relapse than in diagnosis. And now that we know that we also often see it in subclones, Margo will be looking in the next project to...

Charles De Bock (: 09:15

Hmm.

Jan Cools (: 09:30

to this paired diagnosis and relapse samples and try to find out if, is it maybe co-occurring mutations with RAS that define relapse clones or is it something else that RAS even in relapse, so we still don't know. And that's gonna be an interesting question.

Charles De Bock (: 09:46

Excellent, excellent. Yeah, because it reminds me, we did a previous podcast on JMML, so juvenile myelomonocytic leukemia, and they also have obviously a high prevalence of KRAS, and they often also see spontaneous regression of the disease as well, even when there's a RAS mutation. So clearly context is important. I think that's a really important message. One of the nice parts of your study is actually that in terms of the clonal composition is that you inject the diagnostic clones into immune compromised mice.

E: 1099

that were left over from what I understand, usually were the clones with a single mutation. And I might have expected perhaps that the ones that survived were these multiple mutated clones. going back to the question of why is that the case where these perhaps multiple mutated clones don't have the survival advantage? again, perhaps you can comment on this or were you surprised by that finding of the disconnect between the mouse model and the clinical situation?

Margo (: 11:15

Yeah, so I think first what we have to state is that these were very few patients, of course. And most of these patients did not relapse after. So I can at the moment, two out of 13 patients did experience relapse, but so the others didn't. So it's maybe also difficult to look at this MRD cells and then say something about

a risk of relapse as these patients did not. However, think definitely this studying this relapse is, as Jan also said, is very, very interesting as it is important. So we would want to look at para-diagnosis relapse cases and then potentially also look at their MRD samples and look at which clones persist there. And then also look at more PDX samples.

Charles De Bock (: 12:02

Mmm.

Margo (: 12:10

as well with this single cell technique, because maybe what we see in the PDX is not really the same as what we would expect in MRD, but more what we would expect if a relapse would occur. So it's more like the relapsing clones that might come out of it. And this is something we do not know yet at the moment, which is why we definitely want to investigate this in more detail.

And also it's still of course unclear if we can link this PDX mouse models to the humans who are treated with different kinds of chemotherapy. So yeah, think that's also, that remains a big question.

Charles De Bock (: 12:54

be seen.

Yes. mean, an example what you could do is you could treat the PDX mice with the same chemotherapy regimen and then see the MRD cells left in the mouse and then do a pairwise comparison and see what happens there. But I'm guessing that might be for a future study. In fact, I mean, one of the interesting aspects of the study is that you not only looked at the diagnostic sample, but profile those very rare MRD cells, which are so important in terms of risk, intensive risk adapted therapy, because they are

Charles De Bock (: 13:24

directly and still the strongest prognostic factor for relapse. But they're very rare. And at the end of induction or end of consolidation, you're talking 0.1 % or less in some cases. this is cost prohibitive now in terms of the ability to roll this out clinically compared to multiparametric flow cytometry, if cost was not a factor,

Margo (: 13:27

Yes.

Charles De Bock (: 13:48

and we could do single cell analysis of MRD cells. What are some of the advantages and limitations of this technique in terms of what we know about the knowledge and maybe the clinical outcomes for these patients?

Margo (: 14:00

So I think definitely with the single cell sequencing is what you can find is really the different sub clones that are remaining, not only the mutations, but you really know which mutations are still together and which are in different cells. So I think this, for example, what we saw in one patient, I think there was a diagnosis of a big KRAS and a smaller NRAS sub clone. But then at the end of induction, we saw that

mostly KRAS subclone remained and the NRAS subclone was almost disappearing. So of course for this patient after end of consolidation, we saw everything disappearing nicely. for this patient, didn't say anything specifically, but it could be of interest to study this in more patients and see whether this might indeed be indicative of anything. And also I think it's an unbiased

way of looking at the MRD samples with an Amplicon panel, whereas sometimes with MFC or with PCR you have to base on the diagnosis sample and what you find there. So that's also an advantage of the sequencing technique. Of course there are also disadvantages involved and at cost is the major one I would say at the moment.

Also, experience with bioinformatics is necessary at the moment. You need some experience with it, at least to get into the real details of all the single cell data. But I'm a medical doctor and I manage, so it's not like you need a whole bioinformatics degree,

Charles De Bock (: 15:49

maybe perhaps with AI coming on, that's also going to help us do the analysis quicker so we don't have to do the experience here. Sorry to interrupt there, but it just reminded me that AI is such an important part of the future perhaps in helping us do the analysis.

Margo (: 15:54

Definitely possible. Yeah, No, problem.

Yes, and there are also quite easy pipelines. guess definitely with an AI step and the pipelines, you could get to a more easy looking at the clones. But also what you said is that there's very little cells that are remaining at these time points. So it was difficult for us to see sometimes if there were...

Charles De Bock (: 16:17

the answer.

Margo (: 16:26

very little mutations left to see if these were definitely the leukemia cells. If we had multiple cells, for example, four or more cells in a subclone, we could then use the copy number variants to really look if these cells are also high hyperdiploids. And this way we could be quite certain that these remaining mutated cells were definitely leukemia cells. But if we could only find a mutation in, for example, one cell,

It remained a bit difficult to determine, is this really like a leukemia cell? Yes or no. That's sometimes challenging. Yeah.

Charles De Bock (: 17:01

Yep. Challenging. Yes.

I'm guessing with as technology progresses, that will become better. Sorry, Heidi, you wanted to add some...

Heidi Segers (: 17:11

Yeah, I also think we don't know yet if single cell DNA sequencing, which is a possible more sensitive method, has a prognostic advantage above the currently used MRD PCR and flow cytometry for risk stratification in clinical trials. And future possible larger studies are needed for and can provide interesting information in the future.

Charles De Bock (: 17:35

Yeah, because one of the next question I was going to ask is because, MRD is a measure of the fraction or the percentage of cells, but is agnostic to the mutation and clonal heterogeneity. So do you think in the future is it more important to the number of cells which are remaining after end of induction or end of consolidation or more the clonal composition of the MRD cells, which is going to be more important?

for prognosis and that's a probably is a very speculative question, but is that something that we can address maybe the single cell? I'd love to hear your thoughts on that, the number versus the clonal composition.

Margo (: 18:16

So maybe Jan can...

Jan Cools (: 18:18

Okay, I can take that question.

So yeah, I think it's a difficult one, you say, it's speculation. think right now it's mainly the number that you're looking at and we know very well from larger studies that this number is important as a prognostic tool. That's why we need more studies at the single cell level to really understand if the specific mutations there or combinations of mutations, if that even has

Charles De Bock (: 18:24

Mm.

Jan Cools (: 18:48

other types of predictive power or better or worse can also be because potentially if the clones are remaining, we don't need to know exactly which mutations are there. We just need to know that they are there and how many they are there. I think the power of single cell is that you can indeed get that information of that these mutations are there. So we can now address this question, are combinations of mutations predictive for relapse or not? And that would

potentially then allow later on to interfere earlier than the relapse with new treatments, maybe targeted treatments if you know that these RAS mutations are there. More recently, there have been some new RAS inhibitors and even RAS degraders that have been described. So that would allow us to interfere at this earlier time point and try to eradicate these cells further if needed, of course. But that's the question we hope to address in the next study. And that's what we need to know now before.

we can take that into clinical practice. And as Margo answered in the previous question, I think that's one of the limitations of single cell studies is that we always talk about single cell analysis, but at the end, we rarely analyze single cells. We always take five or 10 cells together and that are similar and it's easier still to analyze than one particular single cell. Single cell is...

analysis is noisy. So sometimes we see mutations in normal cells also, and then you can ask the question, if you have one cell with a mutation, is that really a leukemia cell or is that just some noise in the system? So that's another limitation. That's a technical limitation that we currently still have. And it is also something to work on. And that's why here in high-hyperdiploid, it's easy that we have both the abnormal chromosome numbers and some extra mutation. So we have two ways to try to pick up.

the abnormal cells.

Charles De Bock (: 20:40

Yes, it reminds me of, I think, old studies looking at those Guthrie blood spots and then finding that a lot of children had mutations and chromosomal aberrations, but never had leukemia or developed leukemia, which is again, that ultimate question of the ontogeny of the disease.

And when you use chromosome number, think also you added in your single cell analysis the antibody sequencing aspect to investigate the heterogeneity of the cell surface membranes in high-hyperdiploid and that you found that the presence of certain proteins, in this case, CD22, CD69, CD71, CD141 and CD303 could not be linked to a specific genetic subclone. So...

what did the antibody sequencing in single cell DNA, when you combine that, what did that help you do in terms of the technical aspects of your analysis?

Margo (: 21:33

So technically, think it was also very nice to distinguish what Jan said earlier, the normal cells from the leukemia cells. So it was really a nice way to look at these are the normal T cells, CD4, CD8, the NK cells, erythroid cells are also cells that we usually found in bone marrow samples. And so in the beginning, when it was still hard for us,

Because what we see, and I don't think we mentioned this already, what we sometimes see is that only 50 % of the cells had a mutation. For example, one that we picked up with single cell DNA sequencing. But for example, 50 % of the cells did not seem to have a small mutation. And we know from this bone marrow sample that at least 80%, 90 % of the samples are blasts.

at the time of diagnosis. So this was not really what we expected either. And so what we thought is that probably some of the cells already are, for example, high-hyperdiploid leukemia cells, but don't have these additional mutations. so in this way, the protein could help us identify which are the normal cells. And then we could use this copy number technique for which you really need a diploid

population of cells to compare the other populations of DNA subclones that we find too. And this way we could indeed see in some samples there was 40 % of cells that were already high-hyperdiploid but did not have additional mutations yet. And so the protein definitely helped us in identifying these normal cells. In the end we also managed to do this with DNA only.

based on the ploidy of these cells. But it was not always easy to do this with the DNA only. So definitely having the proteins helps to determine these subclones better. What was still an issue with the DNA is that in the beginning, we also thought we found, sorry, with the antibodies, is that we thought we found a new subclone of...

some cells that we could not identify and we thought, this might be interesting because they seem to have also the mutations. But then it seemed that this was actually like an artifact of the protein technique where these cells were probably dead cells that just picked up on antibodies randomly. And so we call them sticky cells now. So they seem to be

Charles De Bock (: 24:10

Hmm.

Margo (: 24:25

The antibodies seem to stick to these dead cells more and so probably they were not really biologically interesting. So that's also something you need to take into account when you look at this data.

Heidi Segers (: 24:36

Thank you.

Charles De Bock (: 24:39

Yes, because now I'm going to drill down a bit on the single cell from a technical perspective because like I said, those cells are very rare and you take a small bone marrow aspirate from these children, which I think is not particularly cellular. so can you just walk me through, because there'll be some listeners to this podcast who will also be thinking about, I'm going to try and isolate this very rare population from a sample, be it an ALL patient or otherwise. And so can you talk...

about some of the technical challenges on isolating these very rare cells, what advice can you give to others who might also be trying to profile a very rare cell population from a primary patient, for example?

Margo (: 25:22

Yes, so it's mostly indeed in the MRD samples that this was important to look for these residual leukemia cells. And so what we first did is we wanted to enrich the samples to have a higher chance of picking these cells up using the single cell technique. And so for this, we used antibodies for CD19, CD34 and CD10.

which are commonly expressed on this B-ALL blast, as we know from immunophenotyping. So we chose those antibodies, we stained the cells, and then we used a magnetic antibody targeting these antibodies. using this magnetic antibody, then we used a gentle sorter, which is called the MARS sorter, it's kind of a nice name for a sorter, from applied cells.

It's really developed for primary cells. It's not a very good way to selectively only have your magnetic cells, but it's really a good way to enrich your cells without putting a lot of stress on the cells. I think that's also important because we don't want a lot of cells dying in a...

sample that is already lacking a lot of cells to begin with. But indeed, it was still a challenge to have enough cells after this enrichment. I think definitely to do single cell DNA and protein, which is also something we tried in some of the samples, you start off with maybe 1 million or 2 million cells if you're lucky in one vial. It's only one vial we have usually of these.

patients at these time points. And then you do a lot of staining steps also with the proteins. So you lose some cells already during all these staining steps. And then after the sorting, you still need to have enough cells to load them in a high enough concentration on the Mission Bio Tapestri, machine is what we used for the single cell sequencing. So you need at least, I think, 200,000 cells.

maybe 100,000 cells in the end to start your single cell sequencing from. So I think with the DNA only part, it was a bit easier to obtain the right number of cells, but definitely that's still tricky.

Charles De Bock (: 28:02

Maybe can I ask if you had 13 samples that you analyzed, were there more that didn't actually go through? Did you start with 20 and you can only analyze 13 because of the sample? Or was it actually, you you had 13 and you had 100 %? Just more of interest.

Margo (: 28:17

So no, so for the diagnosis samples, think all of them just managed. these 13, that was no problem. And then for the end of induction and the end of consolidation samples, I think there was one or two samples that I end up not using the MARS sorter because I already had very limited cells to begin with. So I decided not to do it because that would not have been interesting.

Charles De Bock (: 28:22

went through. Yes, yes.

Margo (: 28:45

And then there was one sample that did not manage to get on the single cell machine. So I lost one.

Charles De Bock (: 28:53

Okay, it's always good to know because people will always see the good results, but people are going, no, no, I did a lot more and these are the only ones that actually came through the pipeline. I'm gonna ask maybe two more questions. So the first one is, there are some studies looking at non-genetic mechanisms that drive MRD. So, at the transcriptional stage, and maybe they go into some sort of quiescence, this concept of drug-tolerant persister cells, which is not particularly linked to mutations.

Charles De Bock (: 29:21

So you you've looked at DNA in the protein and there's obviously one elephant in the room if you will which is the missing part which is transcriptomics. So what are your thoughts on one this concept of drug-tolerant persistors and are you considering doing follow-up studies and adding single-cell transcriptomics to help address this conundrum in the field?

Margo (: 29:43

Yeah, of course, it's very interesting to also look at that. But maybe in the MRD samples, it can be tricky, think, also definitely to look at transcriptomics in a single cell way. So probably looking at diagnosis relapse cases in that case might also be even more interesting to really look at how the transcriptomics change in relapse versus diagnosis.

And yes, definitely, we also are very interested in this and actually a different lab also at the V.I.B. KU Leuven of Professor Jonas Demeulemeester. They recently set up a new technique because what we are very interested in is looking at the single cell DNA combined with RNA in one single cell, because I think that will give the most answers to both what is

DNA and what is a transcriptomic way of relapsing. And so they, at the moment, you still have some techniques that do this, but either they are very low throughputs, they're also still very expensive. But what they did is they set up a way to do kind of a high throughput single cell DNA, whole genome sequencing, together with long reads RNA and ATAC-seq.

So actually it's a very nice way of looking at the single cells. And we also are working together with some of our samples with the hyperdiploid cases that actually experienced relapse. And so we are with their help looking into these samples as well, because indeed it would be very interesting to look at this. So that's something that is coming up.

Charles De Bock (: 31:37

Excellent.

Excellent, because I think that's one of the challenges is that often when we're trying to do that dual transcriptome and DNA, it's plate based and you're only looking at 96 cells and then the question of representation is very challenging. So I look forward to hearing more about a kind of a more high throughput manner to look at both the DNA and the RNA and maybe one day also proteomics, but that's for another podcast, I'm guessing.

Margo (: 31:47

Exactly.

Jan Cools (: 32:02

Yeah.

Charles De Bock (: 32:03

So the very final question I have, it's just when I was reading these papers, and maybe you can answer this for me, high hyperdiploidy is the abbreviation to H-E-H. It just keeps getting, it's like, where did H-E-H come from high hyperdiploidy? I don't know if you've got the answer, but I was reading it going, okay, there must be some sort of history there of why it's H-E-H, because I would have thought it's H-H-Y or something, but.

Margo (: 32:17

Yes.

Charles De Bock (: 32:31

You know, that's my last, it was more just a curious.

Margo (: 32:32

I was also wondering this for a long time because I see it's being used in literature a lot, I think it's because there's also a high hypo deployed and that is shortened H-O-H and this is hyper. So I think that's where it comes from.

Charles De Bock (: 32:46

Okay.

from here. It was more just a curious part because I just always have to, I always like the history of that. So where did that, was it Latin? Did they kind of mix up the words? But okay, very good. I would like to thank you all for your wonderful thoughts on your paper. It's a wonderful paper that's now published in HemaSphere and thank you so much. And I do look forward to the next part of this research project and hopefully it will also be published in HemaSphere. So thank you all very much for your time.

Margo (: 32:51

Thank you.

Jan Cools (: 33:17

Thank you.

Heidi Segers (: 33:17

Thank you.