John DiPersio:

Our session today primarily focused on cellular therapies for acute myelogenous leukemia. And we all acknowledge that there are challenges that need to be overcome using effector cells consisting of either T-cells or other effector cells such as NK cells. So I’m sitting here with Saar Gill, who’s at the University of Pennsylvania and Rizwan Romee who’s at Dana-Farber Cancer Institute. I’d say we have overlapping but distinct perspectives on some of the challenges of treating AML. So maybe, Saar, you could jump in and start off.

Saar Gill:

Sure. Thanks, John. I think that one of the interesting observations that I think we all made on this session is that AML should be an immunoresponsive disease. After all, we all know very well that standard of care for AML, at least for consolidation, is an allogenic stem cell transplant, which is primarily immunotherapy. And yet we’ve all noted through individual ways, including one of our colleagues who’s not here presently, that using different modalities of immunotherapy, be they bispecifics, be they CAR-Ts, that the responses are less than what we expect. And we all have observed again in both as yet unpublished as well as publications that there seems to be a very interesting interaction between the tool, the T-cells essentially, and AML and or the microenvironment, an interesting interaction that may ultimately actually not be to the benefit of the administered immunotherapy and something that we need to resolve. And so, in that context, I was actually very interested that Rizwan has worked on NK cells, which have similarities as well as differences from T-cells with respect to their effective functions.

John DiPersio:

So Rizwan, can you comment on that because NK cells, obviously they’ve become the darling of many around the world who are looking for alternatives for CAR-T therapy, both in the context of toxicities, right? With the limited data out there, there seems to be likely less toxicities in patients, and certainly that’s of interest. Number two is that they work when they’re appropriately redirected to targets. And so they’re effective. They make a lot of gamma interferon, but they’re associated with low toxicities. So I wonder if you could comment on that.

Rizwan Romee:

No, that’s a great question and observation there, which is well acknowledged now through multiple studies that NK-cell and NK-cell CARs don’t seem to cause as much cytokine release syndrome and possibly not cause any neurotoxicity as well. And that may be related to the fact that the cytokine profile, and you showed some elegant data in that regard, that the cytokine profile or otherwise secretome of NK cells seems to be different than T-cells. What was striking from your one unpublished data slide was the IL6 secretion, it was profoundly low or non-existent from the NK-cell CARs, even though the construct was same, target were the same. And yet we know that interferon gamma, NK-cells produce plentiful. In our own observations from the clinical trials, patients who have had up to 80% plus in their bone marrow before undergoing NK-cell therapy, they don’t seem to cause CRS, which is pleasantly surprising. But that may be again related to the fact that the secretome is different. Now that [inaudible] is well established as a safety wise, but what is not well known and is something which we all in the NK-cell field are striving to is persistence. I think T-cells in general, especially the memory T-cell, they can persist for long time. There’s a recent elegant paper from University of Pennsylvania where someone who received CAR-Ts for CLL, almost a decade later there was some CD4 positive CAR-T still persistent and still functional, which is quite dramatic. That kind of data does not exist on the NK-cell side. Now having said that, there’s this recent observation that NK-cells can achieve memory like phenotype. And at least those subsets do seem to persist for longer period of time, but still not years. Maybe one or two months. And I think there’s a big push to enhance the persistence of NK-cells after adoptive transfer. It’s a lot to learn. NK-cells are, in terms of knowledge, in terms of basic science knowledge, maybe a decade earlier compared to the T-cells. I think there’s a lot to learn about NK-cells and that’s an exciting opportunity as well as in terms of intervention, lot to do actually.

John DiPersio:

I mean, one of the things that I’m interested in is sort of the actual ability to expand for NK-cells versus CAR-Ts is less. But I have to say that, look at some examples of diseases that can be effectively treated without a lot of expansion. For instance, well, the few NK-cell patients that were treated with non-hodgkin’s lymphoma, the actual level of expansion in those patients was not nearly what you see with CAR-Ts, but they seemed to have their disease go away. The other group of patients I thought were interesting were the patients that get CTLs. They’re mismatched, largely HLA mismatched. They often are infused in the absence of lymphodepletion, it’s wild. And then over a long period of time, they can eliminate very aggressive B-cell lymphomas. And they have been shown to not really expand very much. But certainly they get to the tumors and they eliminate the tumors. But I don’t know if you have to expand rapidly to do what needs to be done. So that’s why I think there’s also really reason to believe that NK-cells will be a really great alternative going forward. You just have to transduce them efficiently. So what’s been the problem with transduction?

Rizwan Romee:

I think over the years, people have stuck to using, are trying to use VSV-G-pseudotyped lentivirus. And some of our data now clearly shows that that’s not the right choice for transducing NK-cells. There have been several recent elegant papers showing that if pseudotyped lentivirus with ASCT-2 targeting glycoprotein, which the baboon envelope glycoprotein, and there’s also some others, then you can get the lentivirus construct in. Now having said that, there are other groups which are using retroviruses and using different other glycoproteins for serotyping. But in summary, I think in the last couple of years, we have seen that there are ways to get different constructs with lentivirus, retrovirus, some efforts, also sleeping beauty-based transduction methods. And I think we are going to only see expansion of NK-cell CARs and other gene editing because of the advances we have made recently.

John DiPersio:

Any insights into the best endodomains in these cells?

Rizwan Romee:

That’s a great question. There was one study which actually compared in NK-92 cell line, different endodomains. And the winner was the one which had NKG2D transmembrane domain, and 2B4-based intracellular activation domain. Having said that, that particular comparison was made in NK-92 cell lines. And as you know, not everything which we observe in cell lines translates to actual primary cells. In our unpublished data, when we do the comparison, actually 2B4 versus the 4-1BB based endo, it worked equally well. So there is a paper coming out from MIT group where they have done CAR pool approach. It’s pre-published right now. All I can say is that they looked at multiple, multiple endodomains in CAR-Ts and still the 4-1BB and CD28 topped the list, which is interesting.

John DiPersio:

Saar, I think that I can safely say that you’ve demonstrated proof of principle that CAR-T to an AML target works in some patients. So why doesn’t it work in all patients that express the target?

Saar Gill:

I wish I had the answer for that. The things that we are looking at are, I think the obvious ones are, well we have the antigen. Is it expressed at the high levels in the responders as it is in the non-responders? The answer I think is that’s not the problem. The way we’re thinking of it systematically, John, is that patient factors, microenvironment factors, T-cell factors or the CAR construct. And so I think we’ve isolated the CAR as not being responsible. I think with the appropriate preclinical controls, we’ve shown that that works just as well. And so it leaves us with T-cell disease or patient microenvironment. And we’re doing some studies now in a reductionist fashion, which appear to be with preliminary data, appear to be pointing us actually towards the T-cell based lesion, T-cell based if you like problem with respect to that. And so conceivably, patients whose T-cells are better, more fit, younger, have experienced less dysfunction as a result of their AML or its treatment, are the ones who are going to have better responses. So that’s what we’re trying to run down now.

John DiPersio:

So like NK-cells, you think that means that the future of CAR-T therapy for AML is going to be an off-the-shelf approach?

Saar Gill:

Thank you. I think, I’ve said this publicly a number of times, and actually this is the first time that someone has really nailed the sort of inherent logic. If you follow that logic to its conclusion, then the answer should be off-the-shelf, healthy donor. And as you know, there are studies now being done with off-the-shelf, healthy donor T-cells in AML that may give us the answer to that.

John DiPersio:

That’s great. One of the things that I also was struck by is that CRS is a big problem in either bispecific therapy or CAR-T therapy for AML. And we’ve been working on that, but it is an annoying kind of problem that is coupled with low response rates, high CRS rates. So one of the issues is that for bispecifics, you can just stop them. And that’s what you end up doing a lot of the time in many of the bispecific trials what’s happening is that the drug is being given and the patients get severe CRS, it’s stopped. And so then you start it again. And so you’re doing this stop and start all the time, which probably doesn’t provide the best chance for eliminating the disease. So CRS is a major problem. And the other thing is that in addition to that, the response rates are really terribly low, much lower than you would expect. And when you take these same AML cells that are not, you can’t really induce an effect in a patient, and then you take them and inject them into a mouse and then use normal T-cells, you can kill those AML cells just perfectly. So there’s got to be some environmental impact on the T-cell responses which is profound in really preventing optimal responses. So I’m wondering also, we didn’t talk about this, but I’m going to ask you. What is the future of macrophage CARs? Since you published a nice paper in Nature Medicine, I believe.

Saar Gill:

Biotechnology.

John DiPersio:

Nature Biotechnology, looking at using a novel effector cell for the treatment of various cancers. And especially, I think you’re interested in solid tumors. So I wonder if you can talk about that?

Saar Gill:

Yeah, sure. Thanks, John. So I think specifically, the problem with macrophages for myeloid malignancies as you well know, is since the starting cell population is a monocyte, we don’t necessarily know that the monocytes we’re getting from our AML patient, in fact, they’re likely to be, they may well be malignant, particularly AML or myelodysplastic syndrome. So I’m not sure that myeloid malignancies are going to be where I would test the autologous macrophage therapeutics. Having said that, there may be some data from others on utilizing macrophages in myelodysplastic syndrome for example. On the other hand, macrophages have an interest, two very different effector function modalities compared to T-cells or NK-cells, which is they don’t tend to kill by granule exocytosis, but rather by phagocytosis and antigen presentation. And it is the latter that I think is particularly interesting in solid tumors, because these cells will not only gobble up your cancer cell, but also break it down and then present antigens to T-cells and thereby induce a broader immune response.

John DiPersio:

So that cross presentation might be very important as a sort of pseudo vaccination approach. So the other thing that I wanted to point out was with your stuff, Rizwan. So what do you think, are you moving forward with this NPM1 approach?

Rizwan Romee:

Absolutely. I think we have seen promising activity in in vitro and preclinical mouse models. So we are working on a clinical trial where we’ll test this product in patients with relapsed/refractory AML, either before transplant, meaning they failed chemotherapy, or who have relapsed after transplant. I remain optimistic because of two reasons. One, we will use the memory like platform, which on its own, without any gene editing has shown modest to promising at least temporary ability to induce remission in patients with relapsed/refractory AML, including now after transplant. And then if we arm these, so my hypothesis is that if we arm these cells further with a CAR we direct towards the new antigen derived from the NPM1 mutant gene, we should see enhanced efficacy of that product. So we are optimistic about it. I think the biggest hurdle has been viral vector because the viral vector involved is baboon-based. And there aren’t many vendors right now which are making that lentivirus. So we are moving forward slowly and steadily. But yes, absolutely.

John DiPersio:

I like the idea also because we always think about why people relapse with AML. But my question is why are people cured with some kinds of AML? And so obviously you’re considering a cellular therapy in a class of AML, even though they’re relapsed, that traditionally is better and more biologically amenable to therapy. So I think if this is going to work, it might work best in the NPM1-mutated AMLs that you’re focused on. So I’m excited to see how this goes. The other is, obviously we’ve all gone down this rabbit hole and probably I’ve dug one of the deepest holes in the world looking for novel antigens to go after an AML, both using proteomics approaches and genomics approaches. And I’m wondering, Saar, I know you’ve done a little of this. What is your thought? Are there any uncharted territory there? Have we figured everything out and there just aren’t any really differentially expressed antigens on AML?

Saar Gill:

I think we might have both started digging at the same time. I’m surprised we haven’t met in the middle somewhere deep underground, actually. I, for the moment I’ve lost some enthusiasm for doing that. And that’s because I don’t know that there won’t be, you’ve got to start somewhere. And I think it is important to be rational and scientific in you’re thinking. And I think in the individual patient, yes, there might well be some specificity. But I think when you’re developing a drug, that’s not something that’s workable. And so I for one have focused on a different approach, as you know, which is trying to essentially synthesize a leukemia specific antigen using a combination of CRISPR gene editing with CAR-Ts and using the transplant as a platform for doing that. And that’s how I think you can, in fact, credibly you can, this has been of course reproduced by the groups now. You can credibly make a leukemia specific antigen in that fashion.

John DiPersio:

So some of the potential approaches have been using, especially in pediatrics, focusing on some of these rare fusions generating novel antigens. So that’s probably going to maybe yield a beneficial target in a small percentage of pediatric AMLs, but probably not too many adult AMLs. And then the other is just a really broad-based proteomic approach to AML. And looking at not only regular proteins, but embryonic antigens, modified glycosylation, some of the things that we normally think of as being expressed in germ cells like NY-ESO-1 and MAGE-1 and things like that’s, those are probably not the best examples, but there are other examples of probably embryonic antigens that may be slightly differentially expressed in these primitive cells that might be targets as well. I don’t know, what do you guys think?

Rizwan Romee:

I’m hopeful that, yes, interestingly AML has one of the least number of mutations as shown by so many people in some of the finding work led by your group. And yet, it’s hard to treat. We are dealing with very small set of driver mutations, and yet, it’s been the hardest [inaudible] challenge. But having said that, one of the things which excites me about potentially targeting neoepitopes is that if you can find, because most of the cells do maintain ability to express stuff on class I and class II, although in post-transplant setting class II, it does seem to go down. So most of these mutations, potentially, I’m not saying most, but many of these mutations are mutant proteins. There may be neoepitopes, there may be neoantigens, which you could go, but that’s not going to be full panacea or anything. But actually that might sway us towards the TCR-like, or TCR-mimetic or TCR itself.

John DiPersio:

Such a tough problem. And so expensive to dive in. Again, looking at, especially quantitative peptides bound to MHC. This is just a big project, lots of technical expertise, lots of potential issues with efficiency of yielding these peptides and measuring them quantitatively. And then really once you find these peptides, then you have to determine if anybody could ever generate an immune response to them. That’s the other thing. And there’s an interesting group at the University of Pittsburgh who have been developing rapid TCR generation very fast. With any kind of peptide-bound MHC, they have a way of screening libraries for TCRs that bind to them. So within days. And then they can actually subclone the TCRs and express them within a few more days.

John DiPersio:

And so this is potentially a way that people could actually, for every patient, look at their peptide-bound MHC, and then go after potential, not TCRs that are made endogenously, but screening a library of TCRs for ones that bind to these peptide-bound MHCs. And then you could go technically, quickly into the patient. But again, it’s going to be patient per patient, which is going to be something that probably most people won’t want to develop commercially. I don’t know. I think I’d say that based on our session, there are enormous challenges in targeting AML. I think that doesn’t mean that it’s impossible. It just means that we haven’t figured it out yet. And it will take a little bit more time. There’s many ways to skin a cat, and there’s probably a way to do this. And once you find a way, you’ve obviously demonstrated that it can work in some patients.

John DiPersio:

And so there’s just a way to figure out how to do it more efficiently, more effectively. And then I think this is going to be one of the last frontiers in hematologic malignancies, AML. And I think we’ve done a good job with B-cell malignancies, although we don’t succeed all the time. And AML and other diseases are going to be really challenges going forward. I think we’ve made a little bit of progress with some of the other leukemias, CLL, T-cell ALL now there’s a way forward there that is being pursued by some. And I think even in some solid tumors, the data with MAGE-A4 in synovial cell sarcomas looks quite impressive. And so I think progress is being made. But I think that AML, ironically, even though it’s a hematologic malignancy, and a malignancy associated with the fewest neoantigens, will be one of the hardest nuts to crack going forward.

Rizwan Romee:

One other thing I want to comment on, and I briefly mentioned in my talk. We had a patient who had extramedullary relapse and we gave highly purified memory like NK-cell product. And then we did [inaudible], what was striking was the number of T-cells. So there was some pseudo progression, some inflammation, and I was sure he was progressing. Turns out, it was actually NK-cell infiltration, which I was excited about. But when we tried to quantify were they’re more NK-cells or T-cell, actually T-cell infiltration was more there than the actual NK-cells. And they were predominantly CD8 early on, and then they switched to CD4. And that might tell us something, that immune cells are social like us. We like to be friends with each other. And maybe one of the things we should consider moving forward and that applies to both hematologic and non-hematologic malignancies, trying to combine or create constructs that will force crosstalk between or rely more on crosstalk between the immune cells.

John DiPersio:

I think that that is maybe a critical observation going forward. The other thing is some of the things that Marion had reviewed in her talk, which was spectacular, she gave a terrific talk. And I think this notion of targeting multiple antigens is also going to be potentially transformative, and that that could overcome many of the limitations that we find now go going after one antigen. Not only for antigen escape, but the sync issue and specificity of targeting. And so I think the key for that is to figure out how to target two antigens and have the signal occur only when both antigens are engaged, that’s the tricky part of it. Again, there’s a ton of stuff going on there. But that’s going to take a lot of work too to get that more efficient and effective. But I want to thank you all for this session. I really enjoyed it. It was fun hearing everybody talk about AML and some of the advances and challenges going forward. So thanks.

Rizwan Romee:

Thanks John.