Two distinct fetal-type signatures characterize JMML

Charles De Bock (: 00:16

Hello and welcome to the HemaSphere podcast. I'm your host Charles de Bock and today I am joined by Hélène Cavé and Marion Strullu from the Université Paris-Cité and I apologize if I said that incorrectly because today we'll be discussing some of their latest findings in work which is now published in HemaSphere and that is entitled "Two distinct fetal-type signatures characterize juvenile myelomonocytic leukemia." So welcome to you both.

Marion Strullu (: 00:45

Hello.

Hélène Cavé (: 00:47

Hello.

Charles De Bock (: 00:48

I'm going to jump right in and I'm going to ask a couple of questions because there'll be some listeners who aren't completely familiar with juvenile myelomonocytic leukemia or let's just keep it simple, JMML. And so perhaps we can start with how children are diagnosed with JMML and perhaps what their outcomes are in the clinic.

Hélène Cavé (: 00:53

So, that's something that we need to see.

No.

Marion Strullu (: 01:09

Yes, maybe I can answer. So, JMML typically arises in infants and children under the age of two. It is an aggressive disease and most of the time patients arrive with symptoms that are similar with acute leukemia. So, fever, fatigue, bruises, bleeding symptoms and enlarged spleen.

liver, sorry, and lymph nodes. But sometimes, presentation are milder and patient can have only an isolated splenomegaly, for instance. So this makes the diagnosis more tricky in these cases. It's important to note also that some patients have an underlying predisposition condition.

caused by a germline hyperactivation of the Ras-MAP kinase pathway. these conditions are called RASopathy. It includes neurofibromatosis type 1, CBL Syndrome, Noonan syndrome. And so in this patient, they can have symptoms linked to this RASopathy on top of the JMML presentation. And on the...

peripheral blood count. What we can see typically is the hyperleukocytosis with monocytosis, anemia, thrombocytopenia, and also immature precursor cells in the blood and sometimes blast cells also, but not always. This is the typically presentation. Bone marrow examination does not bring many information more than

peripheral blood count. these features are sufficient to evoke the diagnosis. And you asked for the outcome of this patient. And so in terms of outcome, JMML is a quite heterogeneous disease, but most of patients are an aggressive disease and require allogenic stem cell transplant to survive.

Charles De Bock (: 02:58

Outcomes, yes.

Marion Strullu (: 03:12

In our national cohort for patient transplanted, the five year overall survivor (sic) is around 75%. But it is a fascinating disease since some patients will undergo spontaneous remission without any transplant or any chemotherapy. And this is a very intriguing feature of JMML.

Charles De Bock (: 03:34

Mmmhmm.

Yes, I find that actually very interesting and I'm going to ask you about that when we touch on some of your results that you present in your paper because I find that perhaps is there something in your results which might give some insight into which patients may have a larger chance of resolution. And you mentioned, you know, these underlying mutations for the RASopathies where they have hyperactivated RAS pathway mutations. This, I mean, right now we have the molecular era where we have next generation sequencing and that's really providing a greater

understanding of some of the genetic changes that drive all the leukemias. And so for JMML, you hinted at some of the underlying germline predisposition mutations, but what are the other defining mutations for JMML?

Marion Strullu (: 04:20

Yes, in JMML is caused by mutation of the RAS map kinase pathway. This is the hallmark of the disease. And the most frequently mutated genes are PTPN11, NRAS, KRAS, and also less frequently NF1 and CBL, which are associated with the RASopathies I mentioned.

So these are the five main genes mutated in the JMML. And so identifying this mutation are very important to make the diagnosis.

Charles De Bock (: 04:59

Okay, so those would be activating RAS mutations and loss of function PTPN11, all on the same pathway.

Marion Strullu (: 05:05

PTPN11, NRAS and KRAS are activating mutation and CBL and NF1 as they are a negative regulator of the RAS-map kinase pathway, are loss of function mutation or deletion.

Charles De Bock (: 05:10

Okay.

function. Okay,

Thanks Marion for giving us an explanation on the underlying mutations. So how have these genetic lesions impacted diagnosis and prognosis and perhaps patient stratification?

Marion Strullu (: 05:32

Yes, there is not a strict correlation between the driver mutation and the prognosis of patient. Typically in PTPN11 and NF1 driven JMML, patients have an aggressive course and they require transplant to survive. At the other end of the spectrum, CBL driven JMML,

undergo spontaneous remission most of the time. And in the middle, NRAS and KRAS patients have various prognosis. And so for this patient, mostly for this patient, it's important to take into account all the prognostic factors that have been identified. clinical and hematological basic features like the low platelet count.

among the adverse prognostic factors, high fetal hemoglobin level, also overexpression LIN28B, which is linked with the HbF, so low platelet count, and also additional mutations beyond the driver mutation. There are few additional mutations in JMML.

quite special for JMML and pediatric cancer more in general. But in JMML, when they are present, they are associated with a poor prognosis. So this is important to take into account. And also more recently, the DNA hypermethylation, is considered as an adverse prognostic factor. So it helps to take the decision to transplant or not in this patient.

Charles De Bock (: 07:20

okay. And so we're gonna get back to the methylation later on as well because you discussed that in your paper. But in general, what are some of the outstanding questions in the field regarding the biology or the ontology of JMML? And if you want, what inspired your current study?

Hélène Cavé (: 07:40

So maybe I take over? as Marion told, there has been really a lot of progress in understanding this disease from the genetic side and from also the epigenetic spot with the methylation knowledge about JMML. But it's still a very intriguing disease for the reason that Marion told.

for instance, why such an aggressive disease can in some patients be just spontaneously resolved. This is something which of course seems very weird and also because there are very few mutations which are found in JMML and so it raises the question of how can such an aggressive disease happen in such a short time frame in infants with so few mutations and also

how can we explain this short narrow window of onset because actually JMML is characterized by it's really a median onset time of about 18 months which is very very young and it disappears after I mean you cannot have a JMML after a certain age so these are really some conundrums and we hypothesize that it could be related to the

developmental nature of JMML. And this is not, as mentioned by Marion, it's not proper to JMML. It's something which is very now recognized in pediatric cancer in general that some features which are physiological in the ontogeny during a fetal development of hematopoiesis might be subverted or hijacked.

to participate to leukemogenesis, which means that you need very few additional events to have full overt leukemia.

And more specifically in JMML, there has been some recent studies in particular from the group of Elliot Stieglitz in the US, who starting from neonatal blood spots, cards, could evidence that the mutation, the driver, RAS activating mutation is already present at birth, which means that it occurs in utero.

And if it occurred in utero, that means that it occurred in a developmental hematopoietic stem cell. And as we know now, hematopoietic stem cells are very different during the fetal life.

in comparison with after birth in particular because they have some specific transcriptome programs that make them able to meet the requirements which are specific to fetal development which is high self-renewal and a lot of cycling to provide with a very enormous needs of hematopoietic stem cell.

Part of this developmental program, there is really a master regulator, which is LIN28B.

the one that was mentioned by Marion, which is really key to have this cycling and self-renewal. And usually this LIN28B expression is switched off very rapidly after birth, a few weeks after birth. And what is interesting is that there has been a study by a Belgian group of Barbara de Moerloose that evidence in a subset of JMML that was this

kind of an ectopic expression of LIN28B related to the ectopic expression of fetal hemoglobin. So this supported the fact that the mutation is proven to be prenatal at least in a subset of patients and also that in about half of patients there is this ectopic expression of a key oncofetal regulator.

makes it quite, of course, I mean, it supports the idea that there is some some fetal physiology that accounts for the leukemogenesis And that was the basis for our study, which is that we wonder whether the patients who didn't have this LIN28B expression might have some other fetal cues that make them leukemogenic. And that was the starting point of our study.

Charles De Bock (: 11:55

It was intriguing that very narrow onset. mean, we also must say that JMML think, is only one in a million. It's very rare, right, which is probably good thing. But that narrow onset and low leukemic burden, sorry, low mutant mutation burden in the disease is very intriguing. And so I love this developmental concept. And I can see that you did that in your study. In this case, you sorted out granulocyte, monocyte progenitors, common myeloid progenitors, and megakaryocyte erythroid progenitors from both

JMML patients and then compared these to healthy fetal bone marrow and fetal liver samples. I mean, we're gonna get to that technical side in a moment because I find that also very interesting. But in this case, when you compared the transcriptional signatures of all these different cell populations, what did you find and what discriminated those groups when you embarked on this study?

Marion Strullu (: 12:48

Yes, we did an unsupervised analysis, transcriptomic analysis, and we saw that the samples clustered in different groups. The first branching separated LC prenatal, so embryo and fetuses, so prenatal samples from the LC postnatal samples.

This was the two main clusters. And when looking at the JMML patient, we saw that six out of the 16 patients we analyzed clustered with the LC prenatal samples. So we called this group JMML fetal or JMML F. Only two patients were clustering with LC postnatal samples.

and the rest of the patients, so eight out of 16 patients, were clustering within a completely separated group, which was not with the LC samples. And so this group, called it JMML later or JMML-L samples. So when we analyze the features of this

two main groups of JMML fetal and JMML later patient. We saw that in the JMML fetal patient, this patient have low aggressive features as I mentioned, so high level of platelet counts. For instance, we saw more common NRAS and KRAS mutation.

They were also younger compared to the JMML later group, which presented a more aggressive feature with low level of platelet count, age, and all patients with which had PTPN11 mutations were in this group of patients.

Charles De Bock (: 14:52

Okay.

because if we take a deeper dive into those two new clusters that you've defined, that is the fetal and then the later, so JMML F and JMML L what were some of the key transcriptional characteristics that characterized these two subtypes? And you already mentioned LIN28B, so my question, guess, where did that sit and what else did you find that distinguished those two?

Marion Strullu (: 15:22

Yes, we compared differentially expressed genes in those two groups. And first in the JMML fetal patients, we saw in the top upregulated genes in this group, the genes linked to the pyrin inflammasome.

pathway like MEFV gene and also genes involved in the monocytic and dendritic cells like CD14, for example. And so we saw also an enrichment in those pathways by GSEA analysis. was not surprising because

These pathways are very upregulated in the fetal samples. then with ontogeny, it disappeared physiologically. But it's a feature of a fetal sample, physiological samples. What was more striking is what we found in the JMML later group. Because in this group, the...

the most upregulated genes were surprisingly LIN28B and WT1. So it was unexpected because we were expecting that these genes would drive the samples to cluster with the fetal samples, but it was not the case. They clustered completely separately from the fetal samples, whereas this was the master upregulated gene.

those genes are very involved in the fetal program of hematopoietic stem cells and also are dysregulated in many pediatric cancer in general. So this was very striking.

Charles De Bock (: 17:17

Yeah, for me as well, I think when I read the paper, was the fact that the JMML-L had the L28B. it got me thinking that, I think you also mentioned that in the F fetal ones, the spontaneous resolved cases were more in the JMML-F than the L. And so I'm going to challenge you both now and ask, you know, is it, and Hélène, you mentioned this hijacking.

of normal developmental pathways already. So can we talk about that perhaps in the JMML-F? Is it possible then, and perhaps this is very speculative, but that the F JMMLs are hijacking the normal developmental pathways and so can still differentiate and resolve through the normal pathways, and that's why those cases are even in the presence of RAS mutations, whereas the L

they don't have the normal developmental pathways existing anymore, but perhaps they've just ectopically activated this fetal oncogenic protein with the LIN28B so that they know they need only that pathway to transform to leukemia, but then they can't resolve spontaneously because the normal differentiation pathways have now completely like don't exist anymore. Is that something which is crossing?

your minds or am I on the right track? don't know. What are your thoughts on that?

Hélène Cavé (: 18:40

I think it's more than crossing our minds and I think we're very happy that you understood that from the manuscript.

Because exactly the hypothesis is that actually the true fetal signature is captured by the fetal group, which co-clusters with the healthy samples. And it's not driven primarily by LIN28b, but more it evidence something which is, well, not so much

known but a little bit known by people who are experts in the field which is that in in fetal physiology there is a kind of a

priming or at the stage of progenitors of the innate immunity and monocytic markers, which later in life is not primed so early anymore and just appear in mature cells. So this is really the complexity of this physiology that we captured and it's driven by many genes which are

co-expressed and very much in correlation, which I think, it's our hypothesis at least, is consistent with the physiological signature. Whether if you look at JMML, we called them later because we didn't know how to call them because they were just out of anything. So say JMML, JMML or something like that. But if you see them, it's a very...

Hélène Cavé (: 20:12

poor signature in some way because you got LIN28B, you got WT1, but you got very few, I think if I remember well, it's not more than 20 transcripts that were overexpressed in this group. it's very strong, but very schematic in some way. And we think, but that is not really proven.

Hélène Cavé (: 20:35

what we discussed, we think that it's just something which probably is not physiological, but really something which is re-expressed maybe afterwards, and that leads us to the kind of concept of dis-differentiation, which means that LIN28B expression by its own is not able to capture the whole fetal signature, so that do not drive the...

clustering to the fetal samples.

Charles De Bock (: 21:03

So, mean, in that case, all the normal pathways which would regulate LIN28B transcription and post-transcription mechanisms have all been abrogated completely, and so the normal pathways no longer exist in that context anyway, because it's just switched it on just for the purposes of leukemogenesis in this case. And so, if we move on then, and you mentioned this before,

previous groups have shown that methylation is also a key player in JMML. And that was quite distinct. The previous papers have shown that there's hypermethylation and hypomethylation groups. And so how do those distinctive clusters relate to your new definitions for the JMML- F and JMML- L? How do they relate? Do they cluster together or do they spread between the two groups for the methylation status?

Hélène Cavé (: 21:52

Yeah, indeed, as Marion mentioned.

There has been the description of two main groups of well, which can be differentiated according to the methylation stages with one group, is called well, methylation high because there is really over methylation of DNA, which has been described by three groups, the US, Japan and Europe. So it's some very strong finding and also that it's associated with specificities in the patient's presentations, which is older age, PTPN11 mutation.

an excess of additional mutations and also poor prognosis. That's very strong poor prognosis marker. And of course, as Marion told, there was also this kind of trend. Of course, we are dealing with very few patients and heterogeneous disease, so it's not very strong data, but still there was this trend of having in the JMML later, older age, PTPN11 mutations, additional mutations, poor prognosis. So we felt that there might be

relationship with the methylation that was described by other groups. So we analyze the methylation of our patients and we could separate those two main groups of low methylation and high methylation and we compared the transcriptome of those two groups.

Then we get a differential signature and the most differentially expressed genes between the high methylation and the low methylation group was LIN28B and followed very closely by gamma globin, which is of course coding for the fetal hemoglobin. And if we go a little bit more in depth in the signature, we could find some level of overlap.

between the signatures that we define from the later group and the high methylation group. So it's not absolutely 100 % similar, but we feel that our data give a of a biological insight into what defines the difference between high methylation and low methylation. So I think it will have to be proven a bit more on maybe larger.

series of patients, but it's the first time that there is some biology put behind this differential methylation in JMML. And maybe if we can follow up on the last point, what is also interesting is that there is absolutely 100 % correlation between high methylation and LIN28B expression. That is really 100%.

But if we see the control, put the healthy sample controls that we put in our experiments for methylation, what we can see is that those fetal samples, they, of course, as expected, hyper-express LIN28B, because that's really a fetal regulator, but they are not hypermethylated. So what we can just see from those experiments is that...

The link between methylation and LIN28B expression is found in JMML, but not in healthy samples, which goes together with the discussion we had before that the mechanism for LIN28B expression might not be really the physiological one, but driven by some, maybe methylation or whatever, but something which happens afterwards.

Charles De Bock (: 25:15

Excellent. And you also then, you mentioned the fact that you, so your population is small, you've got 16 patients. And in a previous study, which was by the Barbara de Moerloose in Belgium, and they carried out gene expression on 82 JMML patients. And they described at that time, this is 2016, the existence of a fetal-like subtype of JMML. how does...

that particular definition relate to your JMML later or the fetal. So how do we compare and contrast these two different studies, the previous study and your study now?

Hélène Cavé (: 25:54

Yeah, I think that's very interesting to compare the conclusions and the real data from their study. we participated, actually, to their study. we like it very much. But as we told before, first, mean,

Charles De Bock (: 26:08

Hahaha

Hélène Cavé (: 26:16

their conclusion was that there was a fetal-like group of JMML based on the fact that there was the LIN28B expression. And of course, we come with something a little bit different as a conclusion, at least apparently, which is that there is a fetal-like group of JMML which does not overexpress LIN28B. And the group which expressed LIN28B is what we call the kind of pseudo-fetal or something like that.

Hélène Cavé (: 26:43

So it seems to be discrepant, but I don't think it is really. But it gives us also some information about how to conduct studies on this topic. So when we analyze the differences between their studies and ours, I think there are two main differences. The first one is that they started from total nucleated cells for making that comparison. While we...

used some separated sorted populations of progenitors.

So if we go to the fetal signature that we identified, it would not have been possible to identify it from total mononucleated cells from JMML. Because of course then you got a lot of monocytes. So the specific signature is really seen because we separated the progenitors because otherwise, of course, because that's an early priming. So it couldn't be possible to see that from the total bulk of cells. So that's the first point. But the second point for LIN28B is

Hélène Cavé (: 27:43

when I mean the way they conducted their study was really to compare the JMML to a normal postnatal bone marrow and of course what came up first was LIN28B.

But then, they go to the literature. And we know from the group of Connie Eaves or other groups that LIN28B defines a signature that was previously published in mouse, which is the fetal signature. So they apply that and so that there was an enrichment in this signature. And then they concluded on this basis that it was fetal-like. But they never used some fetal samples.

the interest, I think, of our study is that we were courageous enough to take some fetal samples. And then we can really contextualize the LIN28B expression, and we have a more complex signature of what is really a fetal developmental cell. And that is why I think we could make the difference, because we didn't have this kind of a separate signature.

that we could see was enriched because of course we have exactly the same result actually than the Belgium group. It's really similar. But we could see that it is not enough to make cells co-cluster with fetal cells. It's kind of an extra step that makes this difference between what is maybe really fetal and what is just pseudo-fetal or something like that. Don't know if my answer was clear enough.

Charles De Bock (: 28:56

Hmm.

lost.

that was a great refinement of then the previous study and now you've refined it further. And that is actually mostly because you did have access to fetal bone marrow samples and fetal liver samples. I'm gonna, there'll be a lot of other people out there listening who will also be interested in using those samples. So maybe I can ask you some of the challenges you had facing the study in.

obtaining those samples and then also sorting the different cell populations from your JMML from the technical side to ensure you could refine the previous transcription analysis to be in this case more refined at the fetal development stage.

Marion Strullu (: 29:56

Yes, maybe I can answer. And you're right, it was very challenging to do this study and to collect these samples. We set up collaborations with abortion centers, fetal pathologists, and also in the clinical world of the hematology unit pediatric to collect LC bone marrow samples.

But yes, it was very, it was tricky to set up this collaboration, but and the samples are quite rare when we talk about fetus samples. So this was the first challenge, but we are closely implanted in the hospital network. So it helped us to make this collaboration possible.

And the second challenge was the technical part because so you have to be very, very rapid to process the samples because the cells degrading very, very fast. So you have to be always ready to process the samples.

and also very precise to dissect the fetuses and the embryo to maximize the number of cells you can collect. And this was very challenging. Sometimes there were very few cells, viable cells. So after when you sort rare sub-population progenitors and we tried also stem cells, but at the end,

Charles De Bock (: 31:11

Mm.

Marion Strullu (: 31:28

there were very few cells. So we processed a lot, many samples to succeed to collect the samples we analyzed in the study. But the yield was quite low. you have to work hard to be patient, yes.

Charles De Bock (: 31:46

Be patient.

Charles De Bock (: 31:51

I guess I could ask, would you do it again? Or are you still collecting samples, ongoing samples now? guess that's it. Are you going to never want to do this project again because it was too challenging? I guess that's the ultimate question.

Marion Strullu (: 32:07

We try to continue to collect samples because it's still a topic we want to understand. Now that the collaboration has been set up, it's easier. So we try to continue.

Charles De Bock (: 32:30

That's the lessons. Yeah, of course.

Hélène Cavé (: 32:30

I can just have a quick word that after Marion make

all this work of organization, of networking, which was really, really very courageous from her, think, now we want also to consolidate that not only for us, but because we have a lot of requests from other researchers that are working on...

Charles De Bock (: 32:48

Mm.

Hélène Cavé (: 32:51

leukemia. so it's something that we want now to really settle, to be able to help researchers and to be kind of a provider of this kind of sample. So it's of course outside of the scope of the study, but it's something that is really interesting for many people in France and that we really want to organize better to be able to provide

Charles De Bock (: 33:02

Okay.

Hélène Cavé (: 33:14

researchers which are not embedded in a hospital environment to get access to this type of samples.

Charles De Bock (: 33:21

Absolutely.

Absolutely. I think that's a fantastic initiative. just the benefit of knowing that you're out there, you can access those samples, like you said, for researchers which aren't or do not have hospital connections. And I myself have that challenge as well, when you're trying to go from the lab and then see the clinical utility or something, or whether it's relevant to a patient and you don't have relevant patient samples or normal controls.

does prevent some of the broader implications of the research. So that's fantastic that you're giving your resource or allowing people to access your resources or teaching them how to get those resources. That's fantastic. And so maybe through that, like what's the next step? This paper is now published or on the cusp of being published and on HemaSphere. What are the next steps for you and the team on the JMML?

Hélène Cavé (: 34:14

maybe I can just, well, we are not absolutely sure of what we will do, but what is for us something really interesting is to understand the mechanism for this overexpression of LIN28B And a possibility which has been shown in some other models is that it might be directed by an inflammation.

signal through NFkB because there is a direct transcription regulation of LIN28B by NFkB And in other publications that we had with a group from Utrecht, we could see that when we compare this time the PTPN11 mutants,

JMML with normal bone marrow, what comes first as a deregulation is TNF-alpha and inflammatory signal.

And this has been confirmed in the zebrafish model. we're quite sure that there is some inflammatory consequence of PTP11 mutation, which is exactly the group which overexpresses LIN28B mostly. So we now would like to better decipher the link between inflammation and LIN28B expression in general. And for that, probably we'll use some...

cell model and in particular cell models derived from fetal samples to be really in the environment that is the most suitable for a JMML study. yeah, that's what we would like to do.

Charles De Bock (: 35:48

Okay, well, watch the space, I guess, everybody. Well, on that note, I'd like to congratulate you both on your fantastic manuscript, which is going to appear in HemaSphere, and taking the time to share your insights and thoughts on the field and the process which you undertook the study. And thank you, everybody, for listening. And join us next time at the HemaSphere podcast. Thank you.

Hélène Cavé (: 36:13

Thank you very much.

Marion Strullu (: 36:14

Thank you, bye bye