EHA 2019 | Checkpoint Inhibitors in AML

We’ve been doing a number of trials with different immune checkpoint inhibitors and acute myeloid leukemia and myelodysplastic syndrome. We’ve presented this data now a few times and recently published our clinical results with the

We’ve been doing a number of trials with different immune checkpoint inhibitors and acute myeloid leukemia and myelodysplastic syndrome. We’ve presented this data now a few times and recently published our clinical results with the azacitidine and nivolumab, which is a PD-1 inhibitor in relapsed refractory acute myeloid leukemia.

In general, what we’re seeing is that the response rates with the azacitidine/nivolumab combination is about 35%, which is higher than we have seen with other HMA based combinations in the relapse refractory setting on clinical trials we’ve done at MD Anderson in the last few years. So that is encouraging.

But what’s really more interesting is that the overall survival benefit, especially in the early salvage, namely salvage one patients, is quite striking. We see almost 50% of these relapsed refractory AML alive at one year in the salvage one situation without transplant, which compares favorably to historical one-year survival in relapse refractory ML of about 20%. So overall we think that there actually is a durability factor, and we are seeing that a number of these patients achieve what we call stable disease hematological improvement, which doesn’t really fit in the traditional IWG or ELN AML criteria, but with these immune modulating therapies associated with survival benefit.

We’ve been working a lot to try to identify biomarkers, because in the end of the day, if we can identify the population that is most likely to benefit and use these drugs selectively in those patients while we triage the others to the molecular targeted therapies, IDH inhibitors, FLT3 inhibitors, venetoclax based combinations, we have a number of choices now in AML. So how to use these wisely and pick the right patient for the right treatment so we balance the efficacy with the potential toxicity is the key.

So we’ve been doing biomarkers using flow cytometry, and this actually did show us that patients who before treatment had a higher degree of CD3 T-cell infiltration in their bone marrow had a higher chance of achieving response. We had a cutoff of about 14%, and if you had more than 14% T-cells in your bone marrow before treatment, the response rate was about 55%, compared to 25% in those who had low immune T-cell infiltration, which biologically makes a lot of sense and it’s something that has been shown in many, many solid tumor checkpoint inhibitor and bispecific trials.

This is a easy and quickly doable biomarker. So for commercialization, this is probably the kind of biomarker that we will look for as larger studies are looking at these treatments, but we also did a lot more detailed analysis. One of the things that I found very, very interesting biologically was using a cytokine analysis panel based on stimulating cells. So we’ve worked with a company called IsoPlexis, which has developed a single cell, stimulated, multiplex cytokine panel. They look at about 32 different cytokines produced by the T-cells after simulating them with CD3/CD8 stimulation. And they’ve actually published 10 or 15 manuscripts in the last couple of years, mainly in solid tumor, showing that they could predict to a high degree of sensitivity specificity which patients are going to respond to immune checkpoints and recently also to CAR T-cells.

So when we met with this company, we were interested to see if this held true in the AML situation. So we sent them a handful of samples that we knew were responders and a handful that were non-responders and asked them to show us what the cytokine profiles were in the pre-treatment setting and if these were predictive, like they were in the solid tumors. And in fact, we were actually quite surprised to see that the responders had a dramatically higher ,what we call, poly-functional cytokine index score than the non-responders. So they produced many different cytokines at a much higher level, which actually was associated with a much easier chance to respond to checkpoint inhibitors.

And so we think that this actually biologically confirms what we’re seeing is that there are groups of AML that are very sensitive and already have a activated immune system, and in these people you can push the immune system with the immune checkpoint with bispecifics generate a response.

On the other hand, there are patients who we would call immune desert, immune barren, in which case maybe we are wasting our time running risk of toxicity, and these people should go to other treatment. So going forward, we’re going to use this panel in a larger number and hope to publish this data soon.

We’re also doing RNA sequencing, and we actually had a poster here that was done by my fellow looking at data with RNA sequencing on responders/non-responders before treatment and it was actually quite nice. And the RNA sequencing confirmed exactly the flow cytometry as well as the cytokine panel that there were actually interferon gamma associated cytokine profiles that on RNA sequencing predicted for response just like in the cytokines.

So both at the protein level, at the RNA level, and also on flow cytometry, we’re basically seeing that we need to select these people to make this successful, just like we did for FLT3 IDH, and that’s the next step in developing those trials. That’s ongoing.