EBMT 2025 | Harnessing innate NK-cell memory in high-risk AML

About the insight regarding my talk, which I just gave in Florence, I think one of the major challenges with the use of natural killer cells or simply NK cells for immunotherapy has been the short half-life of conventional NK cells. So traditionally, if you were to take NK cells, they last around 10, 12 days, thus they have a natural half-life in the blood. So if you were to adopt and transfer them into patients, that’s how long they’re going to live. And that’s been a very, very big challenge in the field. If you are going to use a cell type which lasts only 10 to 12 days, how do you expect that cell type to lead to sustained antitumor responses? One approach has been using multiple doses of the NK cell infusions, but the problem with that approach is that you often end up needing lymphodepletion before each infusion, but then also that creates a challenge in terms of use of lymphodepletion, which may not be applicable in many patients, you know, use of multiple doses. 

So several years back when I was a postdoctoral fellow in Dr. Todd Finkel’s lab at Washington University, my work focused on can we generate, first of all, can we discover and then can we on purpose generate these memory-like NK cells in humans. And we found that if you were to take conventional NK cells from peripheral blood of anyone, including patients, and expose them to a specific cytokine cocktail, which includes IL-12, IL-15, and IL-18, they differentiate over a period of several days into what we call memory-like NK cells, also called cytokine-induced memory-like NK cells. 

So what’s cool about these memory-like NK cells is that they can live for months in a syngeneic mouse model around 60 to 90 days. And in patients, we have some data that they can live up to as long as eight months, typically again, 60 to 90 days. But in some patients, we found that these cells can live a long time, which is exciting and makes these cells attractive for adoptive immunotherapy, as I mentioned. They help overcome this fact that they can live that long helps overcome one of the major challenges with the use of conventional NK cells, which is their short half-life. 

In addition to that fact, these memory-like NK cells also have better anti-tumor responses. They produce more interferon gamma when they are exposed or co-cultured with tumor cells. And then also they are better metabolically fit. They tend to migrate better into the tumor microenvironment and they resist suppression by TGF beta better than conventional NK cells. So all of these properties make them really attractive for generating CARs or for that matter, even TCR knock in NK cell based products or any other genetic manipulation if you like for adaptive immunotherapy. 

So my lab at Dana-Farber for the last several years has been really focused on generating CARs and TCR mimetic CARs using memory-like NK cells. And during this talk, I mentioned one CAR against NPM1. So NPM1 is the most commonly mutated gene in AML. And as you know, that in AML, there aren’t any good targets. So most of the CARs which have been attempted against AML so far have mostly included CD123, CD33, and things like that. And they are also expressed by normal hematopoietic stem cells, so there’s a risk of myeloablation. 

So my group, in collaboration with a group at MIT, led by Professor Jianzhu Chen, we identified a specific clone of an antibody which binds to the new epitope derived from mutant NPM1. And we call it peptide-centric, also called TCR mimetic, or simply TCR-like CAR, because it acts as a TCR because it binds to the new epitope expressed by HLA class A2. A2 is the most common HLA allele in Caucasians, around 55% of the population. So we successfully generated a CAR, which worked very well, both in vitro as well as in mouse models. So we are hoping to take that CAR to the clinic, because again, AML has been so far very hard to treat with a CAR, and this specific CAR seems to mitigate the possibility of myeloablation, which is a major advantage. 

And then I mentioned in the talk that, yes, in a mouse model, we can cure a lot of tumors, but many times in humans we still fail. And how can we make CAR NK cells more effective, if you will, in terms of both enhancing their antitumor responses and making them live longer and expand better? So my group has been working on the use of IL-12 incorporating into our CARs. Now IL-12 is not a new cytokine by any means in terms of immunotherapy. If you look back, the first trial was done, I believe in 2002 or so, using human recombinant IL-12 systemically in patients, and there were some toxicities and things like that. And then later on, there were groups which actually incorporated IL-12 into CAR T-cells, and those were associated with really, really bad toxicities leading to early termination of the trials and also led to a few deaths in those patients with severe CRS. So IL-12, despite the fact that it’s one of the most activating, most powerful cytokines, has been associated with toxicity. And in general, in the T-cell world, people are now shying away from using IL-12 in their products. 

We, in my lab, we actually compared IL-12 knocked in CAR T-cells versus IL-12 knocked in… when I say knocked in, I mean over-expressing secretory version of IL-12, CAR NK cells head-to-head, at baseline and upon tumor co-culture. And what we found was that the CAR-NK cells, even when they have IL-12 incorporated, they produce way less quantities of key cytokines, which are associated with cytokine-related syndrome, including IL-2, IL-6, GMCSF, and things like that. Suggesting that CAR-NK cells, which are well-known to have better safety profile at baseline than CAR T-cells, but even when you were to knock in a very powerful cytokine like IL-12, they still produce way less quantities of pro-inflammatory cytokines like IL-2 and IL-6, giving us some hope that IL-12 incorporation will be safer in the context of CAR-NK cells. 

And then we also, you know, there are groups which have been using IL-15, for example, for enhancing CAR-NK cells. So in the study, we actually also compared IL-12 secreting CAR-NK cells versus the IL-15 secreting CAR-NK cells. and we found that even though their short-term antitumor responses were similar, however, what IL-12 did better was A, it produced way more interferon gamma, more TNF-alpha, more granzyme B, and better fast ligand expression. All of those things are important for meeting antitumor responses. But then also, we found that they get less exhausted upon tumor re-challenge compared to IL-15 secreting CAR-NK cells. Again, suggesting that IL-12 is a better cytokine to enhance anti-tumor response in CAR-NK cells. 

Also in the talk, I mentioned use of another target called mesothelin in AML. Now, mesothelin has been heavily investigated in solid tumor settings, including pancreatic cancer and ovarian cancer. And there have been trials both using immunotoxins as well as CAR T-cells against those targets in those malignancies. However, it’s less known in the hematological malignancies world. Work led by Professor Soheil Meshinchi’s group at Fred Hutch has shown that certain types of AML, including pediatric AML as well as AML with KMT2A rearrangement, also called AML translocated, as well as core binding factor AML, they tend to have aberrant expression of mesothelin. And because of that phenomenon, we generated a CAR which incorporates secretory IL-12 to make them robust in terms of antitumor function and we used mesothelin as a target. And we found that this CAR really works well, both in vitro and in a mouse model, despite the fact we use a single infusion and a low dose of IL-12 secreting CAR. 

And the last part of the talk, I mentioned use of CD70 as a target. Again, CD70 is a well-known target in AML and other malignancies, including some lymphomas as well as kidney cancer, and my group has ongoing studies in kidney cancer. But then in this particular talk, I mentioned how when we use the same IL-12 secreting CAR against CD70, it works really well. You must hope that that could be the other potential way to make progress in AML. 

In summary, my group is very interested in AML and developing CAR and NK cells, particularly using the platform of this innate memory or also called memory-like NK cells, arming them with CARs, especially CARs which mitigate the risk of causing myeloablation, so going after targets like NPM1, which again is not shared by the normal hematopoietic stem cells, mesothelin and CD70, all of them have this one common thing, which is none of those antigens are expressed by normal hematopoietic stem cells, so there’s less risk of causing myeloablation as you can get a platform for better safety profile and incorporating IL-12 to make them really powerful. That was what the talk was about. I’m very excited about this work. Not only are we learning a lot, but also we are taking these products to the clinic and I’m excited to see that happen and time will tell by how they will perform in patients.

This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.