Andrea Miyahira: Hi, I am Andrea Miyahira here at the Prostate Cancer Foundation. Today I'm joined by Dr. John Murad at the University of Southern California. He will share his latest paper, Solid Tumor CAR T-cells, Engineered with Fusion Proteins Targeting PD-L1 for Localized IL-12 Delivery. This is now available in bioRxiv. Dr. Murad, thanks for joining us.

John P. Murad: Yeah, thank you for having me, Andrea. It's a real pleasure to be here. So yeah, the discussion we have today is based on a paper that we put in preprint with bioRxiv recently. And overall, the goal for us is to improve cellular immunotherapy for multiple solid cancers and importantly prostate cancer. And so, the approach that we took here is to engineer some novel bifunctional immunocytokines to help CAR T cells in that setting.

And so, where do we currently stand in terms of immunotherapy and prostate cancer? Our group led by Dr. Saul Priceman, back when he was at City of Hope and I was there with him, developed a prostate stem cell antigen-targeting CAR T cell or PSCA. And that went into a phase 1 trial over the past couple of years, just published last year in Nature Medicine.

And the design of the trial included dose escalation and inclusion of lymphodepleting chemotherapy. And there was promising PSA drops across the dose levels 1, 2, and 3, which were indicative of bioactivity. But one of the issues was persistence of the response. You could see that even though we were getting reduction in lesions in bone, it wasn't persistent. Over time, we see increases in PSA levels in these patients.

And so, one thing we kept asking is, how can we improve the initial response and also the persistence of response in these patients? And obviously, in the immunotherapy space, people look towards things like checkpoint inhibitors (CPIs). And unfortunately in the context of prostate cancer, they've been relatively unreliable or unpredictable. And so, part of the reason is obviously because of low immune infiltration; cytokines such as CXCL9 or CXCL10 aren't there. We're not drawing in the immune populations. And if you do, you have high levels of PD-L1, PD-1, other checkpoints such as LAG-3, TIM-3, among others, high levels of TGF-β, IL-10 and things like that which contribute to a more immunosuppressive phenotype which limits the effect of checkpoint inhibitors.

And so, we thought about how can we empower both of these combinations of CAR T cells and some of these immune-based therapies? And so we thought, "Let's see if we can incorporate the inclusion of cytokines to help boost both the CAR T cell and maybe synergize with the checkpoint inhibitors." And so we designed these bifunctional fusion proteins which have an anti-PD-L1 scFv bound by an Fc linker, and then either different types of cytokines that we tested, so IL-12, IL-15 and a TGF-β trapping receptor molecule. And this would be co-transduced with a CAR T cell of interest, in this case PSCA, among others. And it would secrete out into the tumor microenvironment and bind PD-L1 in those tumors and see what type of activity we could get both on tumor microenvironment remodeling and immune function.

And so, the first step was to engineer these bifunctional fusions. And so, we set out just to look in vitro, are we getting PD-1 blockade? And we can look among the controls here as the anti-PD-L1 binding relative to the non-binding controls show that we're actually blocking PD-L1. And importantly, on top of that, we're displaying cytokine on the surface of tumors that express PD-L1 simultaneously. So the secretion of these bifunctional fusions aren't only blocking PD-L1, but they're also displaying the cytokine of interest, whether it be TGF-β trap, IL-15 or IL-12 on the surface of the tumor cells in that space.

When we put these CAR T cells in co-culture with tumors in a recursive tumor re-challenge assay, we see that the IL-12 and IL-15 do benefit CAR killing over time. IL-15 drives T cell proliferation, IL-12 to a lesser extent. But the IL-12 drives higher levels of CAR as displayed by our truncated CD19 marker for CAR transduction as well as importantly increases persistence of interferon-γ over all the tumor re-challenge time points, which the other cytokine combinations fail to do over time.

So the question is, do these cytokines and these modifications have any benefit in vivo? And so, we went to our prostate cancer model in a syngeneic mouse, PSCA knock-in. And what we found here is the PTEN KRAS PSCA tumors engrafted subcutaneously treated with these combinations of cytokine and CAR, we see that the IL-12 out-competed all the other groups. We have here, the red and the gray, the gray being a non-binding control and the red being the full anti-PD-L1 IL-12 fusion, had a complete curative response rate. All mice were cured relative to 50% of the non-PD-L1 binding, suggesting that there's synergy between PD-L1 ligation in the tumor microenvironment and the sustained therapeutic response.

Importantly, the non-PD-L1 binding controls, so that being the mutant in the dark gray or a soluble IL-12 that we injected into the mice in combination with our CAR T cells show severe toxicity as displayed by massive weight loss in those groups that resulted in death in a lot of the mice. Whereas the fusion with the PD-L1 binding mitigated these weight loss toxicities. We attribute this weight loss and this toxicity systemically due to increases of interferon-γ found in the serum of these mice. So when you have a non-PD-L1 binding control here, we see a lot higher levels of systemic interferon-γ, which we believe contributes to the systemic toxicity which is mitigated in the context of PD-L1 binding.

So what kind of tumor modifications do we see? Does this IL-12 and anti-PD-L1 localized improve the tumor microenvironment? And so, we ran some digital spatial analysis using NanoString technology, looking specifically at ROIs in the different treatment groups right inside the tumors. And then we quantified in the heat map here to see that we did see dramatic global changes in the immune function in the CAR with the anti-PD-L1 IL-12 relative to the controls.

Furthermore, we quantified this specifically on the lymphoid, myeloid and APC functions, and we see that we have a dramatic increase in CD3, CD4 and CD8 in the IL-12 fusion versus controls. Activation marker CD28 is up as well. In the myeloid populations, we see a drop in suppressive CD14 and CD163 populations suggesting an M1, M2 shift that's favorable for the tumor microenvironment. And importantly, we see increases in CD11c or dendritic cell markers locally as well as MHC class II suggesting that there's better antigen presentation capacity in the full-length fusion in combination with CAR.

So in summary, we've successfully engineered our PSCA CAR T cells to secrete a bifunctional immunocytokine fusion. The IL-12 anti-PD-L1 outperforms other cytokine combinations in this setting, IL-15 and the TGF-β trap. And this synergizes with our CAR T cells to increase anti-tumor efficacy but also remodel the TME. The local sequestration helps the safety profile of a potentially toxic IL-12 stimulation. And early preclinical data suggests that we can also perhaps remove the lymphodepleting chemotherapy that we've been using in our models and clinically with this IL-12, PD-L1 combination, which we couldn't do before.

Future goals obviously is to incorporate a fully human version of these IL-12 fusions, which we've recently designed, for use in our human CAR T cell models and to improve our clinical applications in prostate cancer and beyond.

With that, I'd like to acknowledge all the help that we had from our group here at USC and at City of Hope and of course, Prostate Cancer Foundation for laying the groundwork for a lot of the work that we got to do early on with the PSCA CAR and an extension with the fusion studies that we've now recently put together. I'll take any questions and look forward to discussion.

Andrea Miyahira: So, thank you so much for sharing this with us. So how far in the tissues does the secreted anti-PD-L1, IL-12 fusion travel?

John P. Murad: Yeah, that's a good question. One of the things that we did in some of the other models that weren't necessarily in prostate was we did an ovarian cancer model where we delivered the CAR T cells intraperitoneally with the tumor. And we did some analysis of cytokine in that fluid, in the ascites in the peritoneum. And we saw that with our non-binding controls, we actually didn't see any IL-12 secreting locally into the peritoneum suggesting that once it binds to the tumor, it stays there quite well. And so, we also didn't see any IL-12 in the serum of those mice as well, suggesting that in the bound version, we see pretty localized delivery to the tumor. And in the non-binding, we see release of IL-12 and interferon-γ as we showed earlier, suggesting that you get a pretty localized effect in the tumor. Now, that doesn't rule out the potential for some of those immune cells that express PD-L1 to migrate out of the tumor microenvironment into lymph nodes and other tissues. But the actual molecule itself seems to be pretty well-localized based on our current data.

Andrea Miyahira: Okay, thanks. And do you think that PD-1, PD-L1 is the optimal immune checkpoint to target in prostate cancer or would something like VISTA or B7-H3 be more effective?

John P. Murad: Yeah, that's a good question. There's a lot of literature that suggests that VISTA and B7-H3 are very good targets in prostate cancer, and I think that's true as a single agent. The issue becomes whether or not that synergizes with CAR T cell therapy? And then specifically in the context of these models, we find that the IL-12 drives a lot of high interferon-γ in the tumor microenvironment, and that interferon-γ PD-L1 pathway becomes a point for us to take advantage of. So whether prostate cancer maybe benefits from just a single agent PD-1 or PD-L1 might not be the case. But when you use it in context of the IL-12, it becomes a scaffold as well as an immune checkpoint blockade target.

I think there is definitely room for combination of things like VISTA and B7-H3 in prostate cancer in combination with our therapy. We do have some early data suggesting that VISTA does go up in the tumor microenvironment following the fusion combination with CAR. So, I think there are compensatory mechanisms that could be targeted in prostate cancer using VISTA or B7-H3.

Andrea Miyahira: Okay, thanks. And in general, this was very promising, but these are mouse models. So, how well have mouse models predicted CAR T selectivity and toxicity in humans?

John P. Murad: Yeah, that's the ultimate question. As I mentioned, we use a PSCA knock-in transgenic mouse in C57BL/6 backgrounds. So anywhere we have PSCA expression, it's going to be also expressed in normal tissues. And so, we can look for some of the off-tumor on-target toxicities. And fortunately, we haven't seen any of those in this model. That's not to say that it doesn't happen and we have to do dose escalations and see if we have targeting off the tumor that might be representative of toxicities that we see in the clinic. So we have to do those studies.

In the context of mouse models such as PSMA, which is a very hot topic and target in prostate cancer, we know that there's a transgenic knock-in PSMA model. Others, Nour Shoubaki at UPenn, among others, have used and shown that there are similar toxicities such as macrophage activation syndrome that have been found in those models that also mimic what's found in humans. And so, we think there are correlations, but it depends on the models and designing those models appropriately so that you can look at those toxicities.

Andrea Miyahira: Okay, well thank you again for sharing this with us, and good luck on the next steps of these studies.

John P. Murad: Appreciate it. Thank you again to UroToday and yourself for having me discuss this today.