Compugen Ltd. (CGEN) Earnings Call Transcript & Summary

Good afternoon, and welcome to the Compugen Virtual Investor Event. [Operator Instructions] At this time, I would like to turn the call over to your host, Anat Cohen-Dayag, President and Chief Executive Officer of Compugen. Please go ahead.

Good morning and good afternoon, everyone. We're delighted to welcome you to our special call today with the man that likely needs no introduction. Professor Drew Pardoll. Drew is the Chair of our Scientific Advisory Board and a world expert on the DNAM axis. Drew's research had the major contribution to the development of the first anti-PD-1 therapy which made a breakthrough in cancer treatment. No doubt, this class of drugs is transforming patient lives and it represents a greater than $35 billion market today, but there is so much more that needs to be done for the majority of patients that are still not responding. For those that are suffering and are resistant to these immunotherapy and Drew is heavily involved in trying to identify mechanisms that are associated with resistance like we do at Compugen. And Drew is a world expert in the possible role of the DNAM axis addressing resistance. As you know, Compugen's hypothesis behind the DNAM axis role suggests that in many patients and in many tumor types, all 3 pathways of DNAM axis, PD-1, TIGIT and PVRIG may be needed to be blocked in order to maximize the immune system response to fight cancer. So before we hand over to Drew, Eran Ophir, who is leading our research and drug discovery, will set the scene explaining our DNAM axis hypothesis and how we're quickly advancing it from a hypothesis to reality as reflected in the clinical data we've presented with a combination of [indiscernible] our potentially first-in-class anti-TIGIT, anti-PVRIG, COM902 our potentially best-in-class anti-TIGIT and PD-1 inhibition in patients typically not responding to immunotherapy. Following the presentations, our Chief Medical Officer, Henry Adewoye, and I will join Eran and Drew for the Q&A. Next slide is our usual forward-looking statement. And now over to you, Eran.

Thanks, Anat. So Compugen is using its pioneer discovery platform to identify new targets and new MOA to address immunotherapy resistance [indiscernible] cancer. And our platform is validated because we already took a few targets from pure computational prediction all the way to Paclical POC. And as you will see today, initial signs of clinical activity. So the 3 major assets in our pipeline right now COM701, COM902 and COM503. COM701 and COM902 which will be the focus of today's discussion, which are blocking the DNAM-1 axis. So COM701 is a potential first-in-class anti-PVRIG antibody. COM902 is a potential best-in-class anti-TIGIT antibody and I will not discuss comparable 3 today, but definitely 2 words is really novel target coming from our discovery platform, another approach to [indiscernible] to treat cancer. But again, today, we're going to focus on the DNAM-1 axis. So we are starting this pathway promote a decade now. We identified TIGIT at the same time Genetec did. A few years later, identified PVRIG that had no literature around it by the time we discovered it. We're the first to take PVRIG into the clinic and following our understanding of the DNAM-1 axis, we decided to develop then COM902 to be combined with PVRIG blockade because we understood the synergizing potential. We generated COM902 with a non-Fc effector function because we believe following our understanding of the pathway that you actually want to reinvigorate CD thesis without the risk of repeating those. And this property of COM902 in combined with a superior binding and affinity and function position COM902 is a potential best-in-class and this is actually validated by the choice of AstraZeneca to select COM902 as the TIGIT arm of their PD-1 TIGIT bispecific antibody and they're also chosen an Fc effector binding for their antibody, and the recently announced that in addition to the multiple Phase II they're having for this molecule, they're going to advance into Phase III later this year. So these are our main wholly owned assets, and now let's deep dive a bit more into the DNAM-1 axis. So PVRIG and TIGITs are both expressed on T and NK cells, and there are negative checkpoints. Meaning that upon biding their ligands, they deliver inhibitor signal inhibiting the T and NK cells from attacking the tumor. And the part of DNAM-1 axis because actually they have a shared ligand, both of their ligands are binding a shared molecule, which is DNAM-1, which is a co-stimulatory molecule, which actually is giving positive signal to T and NK cells to attack the tumor. So people with TIGIT actually hijack these ligands of DNAM-1 and actually, not only by themselves, deliver inhibitory signal, but also prevent activity signal by DNAM. In addition, PD-1 intercellularly -- DNAM and actually also prevents activity. So what we understood that if you triple blockade all these 3 components of the DNAM-1 axis, which is PD-1 PVRIG and TIGIT, and this what we saw preclinically, and we start to see clinically then we get the most optimal T-cell activation and antitumor effects. So while PVRIG and TIGIT are parallel pathways in the DNAM-1 axis and they share some similarities, they're also very different. PVRIG, the ligand of PVRIG is more dominant to certain tumor types like ovarian cancer. While TIGIT is most dominant on exhausted T cells and regulatory T cells, actually, PVRIG is not so highly expressed on regulatory T cells is muscle dominant on stem-like memory T cells, the cells which have very strong proliferative capacity, and we'll talk about the importance of this for PVRIG biology in the next slide. And exactly in correlation to that, the ligand, again, PVRIG/PVRL2 is very dominant on dendritic cells, the same cells that interacts with TSCM to stimulate their proliferative. So most checkpoints are expressed on these T cells exhausted or effector cells in the tumor microenvironment, directly interrupting with the tumor. And we know from TIGIT studies until now at least, that the most activity for TIGIT blockade combination with PD-1 is seen when we have sufficient amount of T cells in the tumor environment. But if you're looking into patients with immune desert or just PD-L1 low indication, which have less T cells inside to begin with, then most checkpoints will be less active and specifically until what we sent in the PD-1 TIGIT combination is not sufficiently potent to drive immune activation and anti-cancer therapeutics and aesthetics. While PVRIG is also active in this stage of T tumor cell interaction, actually the dominant expression of PVRIG, but that is different from all the data checkpoints is on this fan light memory T cells, which have a very strong proliferative capacity. So our understanding now of the PVRIG biology that if you use COM701 to block this interruption of T cells and dendritic cells, you have the potential to additional waves of effector cells coming out from the [ lymph node and TLS ], and now making less inflamed tumors more inflamed, now to combine TIGIT PD-1, PVRIG blockade actually cannot synthesize the tumor to respond to PD-1 and TIGIT blockade and only have a full-fledged TIGIT blockade activating the T cells while they are already a tumor environment. So they say that the proof is in the pudding. So this is exactly what we start to see in our clinical trials from monotherapy to combination with PD-1 to triple combination. You see this activity in places where you wouldn't expect checkpoints to work with activity in indications, which are non-inflamed. We see activity in patients which have immune desert tumors, we see activity in patients who fail prior checkpoint treatment. No one not responding to additional checkpoint. We see, for example, for the monotherapy by the patient with ovarian cancer that has immune deserts before treatment and that patient responded clinically to monotherapy and stayed from only 18 months on the study with deep response, and then we see all of these and we cover some of it today of responses in platinum-resistant, in MSS-CRC, and again, in patients who failed prior checkpoint treatment. So if we look in ovarian cancer, platinum-resistant ovarian cancer is a very difficult indication. These patients have -- the standard of care is a toxic chemotherapy with over response rate of 12%. The option of immunotherapy for these patients is quite poor, PD-1, even in combined with TIGIT, actually give less than 10% of response rate and specifically 0% in PD-L1 negative tumors. And if you focus, especially on the triple blockade when you use COM701 plus nivolumab plus the BMS TIGIT that we use these days, we can really see 45% disease control rates and 4 patients in the time we reported, all of them were ongoing at the time we reported with deep responses and ongoing responses, which is really supporting the activity of the triple blockade in this difficult-to-treat and checkpoint non-responsive indication. So for example, this patient -- this patient, she had 7 prior lines of therapy. She actually received a lot of these lines nivolumab, and she progressed on Ebola. She has no responsive stabilization on nivolumab and then she came to the treatment of the doublet constant nivolumab, you can really see here the partial response that later on even deepens and again, talking about the PVRIG biology driving T cells into less inflamed tumor types, you can really see the increase in T cells on treatment following this doublet combination study. And then the signal that we see in MSS-CRC and even more difficult indication, especially for checkpoints, but even the standard of care in the last line of patients that's getting into these trials, is over response rate of 1% to 2%. So there's a huge unmet need here. And 70% of these patients has liver metastasis and [indiscernible] even bigger. And while checkpoints are very close to 0 response rate in general in microsatellite stable [ sclerotia ] cancer in the patients with liver metastasis, their response is really. I mean overall, even in combination with 0% response rate. In our study, we had 17 patients with the liver metastasis in MSS-CRC. In 2 of those, we had partial responses, which gives an over-response rate of 12%, which is encouraging compared to any type of historical data that we have seen. And what is interesting is, and we'll not go through all the details here, but for example, that patient, had immune desert before treatment and again, MSS-CRC biopsy from the liver metastasis, you can see on treatment, following treatment of COM701 plus nivolumab, this is not even the triplet yet. So we are going to start the triplet soon. But the doublet combination really gave a huge impact of T cells. And what's interesting to see is really the fingerprints of the PVRIG biology of this hardly any thesis before treatment and the massive increase of T cell proliferative in the tumor microenvironment and coming probably from the lymph node from inside the 2 macroenvironment, all of this in this indication that normally would not respond to any checkpoint especially in patient with liver metastases. So before handing over to Drew, so what is next for Compugen. So we are focusing now on these 2 studies. In addition, of course, to AstraZeneca that they are progressing with there bispecific and going to start to pretty soon with the PD-1 antibody based on COM902. We are focusing now on 2 triplet studies combining COM701 with COM902 and Pembrolizumab, and based on the biology of PVRIG and based on the previous data, the some of it you have just seen and the special data are supportive of a potential for a chemotherapy-free option for these difficult indications and with a very favorable safety profile and what we start to see like the really responses and T-cell infiltration in places where normally other checkpoints will not work, and the unmet need here is very clear. And we really think that this might be the right combination to transform patient lives in this indication. And by that, I will hand over to Drew.

Great. So Eran, can you hand over the screen to me.

You'll need to share your screen.

Yes. So it's saying that -- I'm viewing a wrong screen and it's not giving me the option to share screen.

Okay. Eran, you can end your screen share and then Drew, you can try it.

So my screen sharing is not -- there we go. Great. Thanks so much, Eran, for that great introduction, and thank you so much, not for that extremely embarrassing introduction of me. Just for disclosures, our group has received research support from Compugen for the PVRIG and TIGIT-related scientific research. This has been a 10-year collaboration, which has been extremely productive. And I'll say at the outset that I'm certainly an enthusiastic based on the data both preclinical and also the early clinical data that this is a very promising opportunity as Anat mentioned, share Compugen's SAB. I have no intellectual property relevant to this presentation nor any personal equity in Compugen. What I hope to do just with a few of these bullet point slides is to give you a little bit of a sense of my take on not only the pathway, but also to address some of the questions that are out there in the community. And this will hopefully simulate questions from some of the folks on the Zoom. So really, I've been a strong advocate to Compugen and I'm very happy that Compugen has really come to focus on really this triplet. For a number of scientific reasons, I think PD-1 blockade is foundational. So I think it's important that, that's one piece of this. And then also, as Eran mentioned, PVRIG and TIGIT really are parallel pathways and so while blocking one or the other may very well add some activity if you only block one that leaves the other checkpoint intact. So it really makes scientific sense to block these concordantly. There, as Eran mentioned, differentially but preferentially expressed really on the 2 populations that we know of sort of precursor populations of tumor-specific T cells that can be when properly unleashed and unblocked to become activated antitumor effector cells. I'll show you just on the next slide, one piece of data from my colleague here at Hopkins [indiscernible] showing that knockout of both PVRIG and TIGIT in animal models have additive and in some cases, synergistic antitumor activity in multiple aggressive mirroring tumors that are resistant to anti-PD-1. Does that guarantee success in the human? No. But if you look at the successful immunotherapies of all sorts, they all started by demonstrating activity in challenging poorly immunogenic murine models. And then also oftentimes not appreciated even above and thinking about translation of mouse data to the human, there are actually 3 reasons to predict that PVRIG will play actually even a larger relative role in humans than in mice. It's expressed more highly in humans than in mice. It's also developed a full item in the human that it actually doesn't have in the mouse, which is the conventional strong inhibitory signaling motif that binds particular phosphatases and also human tumors have some of them, including ovarian cancer, a very high expression of the ligand for PVRIG. So the top just is a reiteration of the fact that there are these 2 populations, stem memory and exhausted T cells. Stem memory, particularly have high level PVRIG lower and exhausted, but where you have higher levels of TIGIT and PD-1, which upon multi-checkpoint blockade, essentially can activate cells from these 2 preexisting tumor-specific populations. These are just -- this is B16-F10 a very classic, highly non-immunogenic mouse model, looking at the various knockouts, and you can see either looking at tumor growth or survival that the combination of knockout of , which is the genetic form of blockade, of PVRIG and TIGIT generates stronger antitumor responses than knocking either of those out alone, which do certainly have activity. Similarly, the combination knockout has a longer survival in an extremely aggressive ovarian cancer model in human. This is the IVA model expressing the GF. So again, I think all of the preclinical data strongly supports this as a promising triplet, but I think it's worth just putting on the table a couple of things that are sort of out there in the community. The first is has TIGIT been a clinical failure? And the answer is we don't yet know. There was clearly in the early rollouts, irrational exuberance about TIGIT and that was, in turn, followed by what I view as irrational skepticism. Just to point to a few of the trials that obviously everybody has looked at, obviously, the Roche trial and important to point out that for the Phase III, the PFS curves have not been reported out. OS has not yet read out. And I think people assume that the endpoint is improvement over atezo alone, but actually, it's not clear that, that is the endpoint because for -- if you're a pharmaceutical company like Roche, the endpoint is not a positive trial. The endpoint is they need to have a result with atezo, which with all due respect to Roche Genentech is not as potent to PD-1 pathway blocker as pembro. But what they need to beat obviously is pembro alone for PD-L1 greater than 50% non-small cell lung cancer. So we don't know that the trial, if the data is ever released isn't going to show an improvement relative to atezo, which would demonstrate some additive benefit of TIGIT blockade. There was also a report of a failure in pembro chemo resistant non-small cell lung cancer, but that obviously doesn't say anything about what the first-line activity would be or what the activity would be in IO-naive patients. And then the Arcus trial, which I actually thought again was those relatively small numbers of patients, encouraging for demonstrating an improvement of the combination over their anti-PD-1, but the fact that the anti-PD-1 alone arm so-called underperformed in my mind, does not negate the value of the improvement relative to anti-PD-1 alone, this could be that their anti-PD-1 is less active than pembro, just like atezo. It may or may not. I just -- I don't think anybody's actually compare them head to head. And also I think it's very possible that there may be patient selection differences relative to the Merck approval trials, which could explain it so-called underperformance. On the right side, the question of does anti-TIGIT need to be "Fc active', namely have an Fc region that binds Fc receptors. The answer is it's actually unclear. I think likely not the case. There have been conflicting reports in mouse models. The first report actually from Vijay Kuroda and Anderson in the mouse reported synergy between anti-PD-L1 and an Fc inactive anti-TIGIT. Clearly, if you needed the Fc receptor dependent effect, then you wouldn't expect to see an effect in TIGIT knockout mice. And I just showed you that you do actually see an antitumor effect in TIGIT knockout mice. I might also remind you that with anti-CTLA4, there in mice, the anti-CTLA4 antibodies do have to be Fc active for the antibodies to have an effect. But that actually is known to not translate in human because essentially ipi has the same activity as Temi. Temi is Fc inactive. Ipi does bind to Fc receptors as a human IgG1. And there's absolutely no difference in their activity whatsoever. Fc active anti-TIGIT also concerns me because the effector cells also express TIGIT, because you're -- and you're trying to block TIGIT. So what if you block TIGIT but also eliminate the very effector cells that you're trying to enhance their activity. So I think they're real, personally concerns with regard to so-called Fc-active antibody, and the reality is, if you listen to the explanations as to why you would need the Fc activity with all due respect to some great scientists, it's all hand-waving. And then finally, why do -- I think these particular trials that Compugen has focused in on are really strong trial designs. Firstly, as I've mentioned, the triplet prioritizes potency with 3 really excellent monoclonal antibodies. And I believe that in drug development in cancer the 3 most important things are potency, potency, potency. Ovarian cancer has always been known to have in a large proportion, a decent lymphocyte infiltrate, yet it has a very low response to anti-PD-1. So it's a high unmet need. Generally, the response rate is less than 10% in PD-L1 negative patients. I know at least one of the responders in the triplet trial of Compugen's is, in fact, PD-L1 negative in the Merck study with pembro alone, there was a 0% response rate in the PD-L1 negative ovarian cancers. Ovarian has a very high expression of PVRL2, the major PVRIG ligand and also some great opportunities for identification of biomarkers for further patient selection. Similarly, MSS colorectal cancer high unmet medical need, very, very low response to anti-PD-1 alone. You saw even the doublet had very small numbers, but had some activity. Also, with liver metastases, there -- I think there are some good opportunities for on-treatment biopsies, which I think is going to give a lot of information that will come out of the trial beyond simply clinical endpoints. And also, again, there will be opportunities for identification of biomarkers for patient selection that will be very important. So with that, I will stop share and turn it over to Sarah to moderate the Q&A session.

Wonderful. Thank you, Drew. [Operator Instructions] So with that, we'll take our first question from Mark Breidenbach at OppCo.

Just a couple for me, probably both Eran and Dr. Pardoll, I guess I'm wondering if TIGIT and PVRIG are somewhat redundant checkpoints that maybe can compensate for one another. Why should we be expecting a doublet, PD-1 plus TIGIT antibody to have any benefit over PD-1 at all? What's kind of the rationale there? And maybe just what's the Occam's razor explanation for the divergence of clinical data we've seen from across the clinical trial space. I know that might be a hard one to answer. And I have one more follow-up after you -- after you fill those questions.

Yes. So -- if you're trying to activate a pathway, and you have 2 redundant upstream targets that you're trying to activate. There redundancy does not buy you an advantage to hit both because if you hit one, you activate it, if you hit the other you activate. If you're trying to block a pathway, then redundancy arguably requires it you block both. Just imagine you have a big -- a bunch of steer in a rodeo or something and there are 2 doors and you want to keep the steer inside. And you have 2 doors into the rodeo space. If one door is slam shut but the other door is open, even though it's a smaller door, the steer can get out. If you want to keep the steer in the pen, you have to shut both doors. So that's sort of the -- and similarly, here, they're not -- they're not totally redundant, but they are somewhat overlapping. And again, so if you're a tumor and you want to block the immune -- the patient's immune system from getting at you and you have 2 molecules, 2 receptors that will each send that are similar related that will each send a negative signal to turn off that T cell. And you're making ligands that will hit both of those receptors. And now you come in with an antibody to block one of them. The other one is still open to be engaged by that tumor. So that's the fundamental reason why if you're trying to inhibit as of to agonize, it's important with redundancy to block both. So to your second question about why -- about the early trials looking like they're "all over the place". If you've seen one cancer patient, you've seen one cancer patient. There, you have to really go into the populations of patients that not only just that are in the inclusion, exclusion criteria, but really look at the characteristics of the patients that went on to a given trial. You need to look at the setting, you need to look at the antibodies. There are a lot of variables, and I apologize, I work in intercity medical center, if you hear sirens periodically. And so you're going to have variability. And this is why the big randomized, properly designed Phase III approval trials are the trials that you need to ultimately answer. Question is, are we seeing signal? And for first-line non-small cell, certainly, what I look for with the TIGIT is signal above baseline comparator, which is anti-PD-1 or in the case of Roche anti-PDL1 alone with the Arcus trial, again, there was a clear signal by adding their anti-TIGIT, which is an Fc inactive. And in the Roche trial, we haven't seen the data on their Phase III trial, PFS, they've essentially announced that they didn't hit their endpoints for PFS. We haven't heard about OS. But I'm not -- it's not clear to me what the actual -- what their endpoint was. Yes, it's an improvement relative to atezo alone, but how much improvement are they asking for? If they're atezo alone as a PFS that's 10%, 15%, which I imagine it will be because it's not as good a PD-1 pathway blocker as pembro. If it's 10% to 15% below pembro even on the Phase III and adding their anti-TIGIT brings it up to the level of pembro alone in patients that are PD-L1 greater than 50%. That may be a positive trial, but that's not the endpoint that matters to Roche. They need a strong enough result that not only do they get FDA approval, but an oncologist says, this is so compelling that I'm not going to any longer prescribe to my first-line lung cancer patient, what I've been prescribing for the last 7 years, which works pretty well, which is pembro because I'm not convinced that their data compared to pembro, it's not going to get me to change my prescribing habits. That's really the operational endpoint, if you're Roche that you need. It's not another New England Journal of Medicine paper. That's good for us as academics. But -- so I think -- again, those are some of the points that I would make to keep in mind before just sort of writing off something based on a lot of really small studies in terms of what's really been reported to the level that you can look carefully at what the variables patient populations and outcomes are sort of under -- below the level of the press release.

And then on the design of the 2 triplet studies that Compugen is running right now, we're testing 3 agents in tumor types where there's not a lot of single-agent activity for any of the 3 components. I'm wondering, I would love to get your thoughts on potentially combining with chemotherapy in these maybe second or third line settings. Wouldn't that potentially be easier from a regulatory perspective than going after a triple combination as the company is doing now.

Yes. So I -- in ovarian cancer, we don't really know whether the chemotherapy agents that are conventionally used in ovarian cancer, how they are going to play with immunotherapy. For reasons that we're actually working hard to understand. There's no question that the platinum doublet, particularly a platinum agent with pemetrexed in lung -- non-small cell lung cancer does have an additive, if not potentially synergistic effect with anti-PD-1. That's not the case with most of the other even FDA-approved combinations of anti-PD-1 with other chemotherapy agents. The chemotherapy agents are so diverse in terms of what they do, the targets they hit, antifolate and terculators, microtubule agents, [indiscernible] inhibitors. So might it be an easier regulatory path, maybe, but there's no data that it's going to be the most effective. I'm a strong believer that the fastest pathway to market is the pathway that brings the greatest efficacy on behalf of patients. I really do believe that the market and the patient-focused incentives are concordant here. So I think that if the data from the -- certainly, from the ovarian holds up 4 out of 20 patients, and I've been lobbying to really pay attention to effects in the PD-L1 negative variance, where essentially pembro has a 0 percent response rate, that results that are reproduced the results in the small 20-patient triplet are going to turn into a fairly rapid, it may be a multi-arm trial, but I think, a fairly rapid pathway to approval. Similarly, MSS, it's the -- the bar is so low. There's just really so little. And again, we don't -- right now, we don't know what throwing these checkpoint -- this checkpoint blocker combination together with sort of standard chemotherapy is going to do. And I hate sort of closing -- just closing my eyes and hoping that I hit the bull's eye.

And the next question comes from Tony Butler at EF Hutton.

My question for you and maybe Eran is this. You can imagine that a tumor microenvironment, which has been -- which is refractory to previous therapy would produce different cytokines or maybe little cytokines, then one in which you do have an antitumor response. So the question is, when you add the triplet, is there evidence that actually there is a change in those cytokines? For example, is there an increase in, say, IL-2 versus that pre-triplet TME? And moreover, the second part of the question might be, are there phenotypic changes in those T cells based upon the pathway being blocked by the triplet versus one in which it's not blocked. And you might also consider back to the previous question, what might happen if it's a not to triplet, but 2 of the 3 components is there any evidence, cytokine or phenotypically for that outcome?

So I will say before handing it over to Eran, who's been very much, obviously, on top of the analysis of on-treatment effects and Eran showed a beautiful slide of one example of changes with therapy. What I can tell you is that in the mouse models and granted, you can't take mouse models as absolute, but the biology of molecules in this pathway looks very similar between mouse and human. So I think in terms of understanding what effects the pathways have in mouse, I think that is going to be at least partially translatable without taking the time to go into the detail, but my colleague, Dr. Ganguly is writing this up now. He's been able to, in the mouse models, compare with the individual versus the combined knockouts and which is really the ultimate clean genetic methodology to answer your question, at least in mouse models, which are in bread and more consistent, that there are absolutely differences in terms of the qualitative and quantitative differences in tumor-specific T cells that enter into the tumor without question. But -- but why don't I hand this over to Eran to maybe say a little bit more about some of the information that's been gleaned from the clinical patients.

Thanks, Drew. So about your question on patient failed prior IO. So I think it's really dependent on why they fail. I'm not sure if the triple blockade could rescue all of them. But for example, if we didn't have enough pieces to begin with, and that's why we didn't respond, then again, giving us also the unique biology [indiscernible] then potentially triple blockade can more importantly increase these in numbers and then induce also clinical response. If, for example, they lost [indiscernible], for example, PVRIG and TIGIT also NK checkpoints, while we don't have a lot of antihuman to macroenvironment, there are some indications for NK or maybe even gamma Delta, which also carry PVRIG and TIGIT that can control some responses even when the MSC is lost. And about the phenotypic changes for interpret blockade, so as Drew just mentioned, is probably easier to follow very precisely in the mice, but also in human subject, what we have shown that following the treatment, you see increase [ inter PhoGamma ] in the blood and most importantly, [ inter PhoGamma ] signature in the tumor, actually, we have some data 6 reduction in responding patients. So definitely, the triple blockade is being able to modulate tumor microenvironment really changing not only the cytokine escape, but also increasing T cells activated polyfunctional T cells. So this is what we see in patients and again, in more details in Drew studies.

May I have a follow-up if possible. But Dr. Pardoll, back to your comments in mouse models, can you say that there were changes in the IV8 model with respect to cytokines pre and post?

Yes. So -- we've -- so the answer is yes. The analyses actually that has been done and is actually ongoing is an analysis in called single-cell RNA seq, which gives you the whole transcriptome of each cell, and you can analyze 5,000 cells per tumor. And so you get a massive amount of information, and there are actually multiple pro-inflammatory cytokines that do go up being made by the T cells. And we haven't even started looking at non-T cells in the tumor microenvironment, also receptors on the T cells for cytokines that may be produced at low levels in the tumor milieu, such as IL-7, the IL-7 receptor levels go up. So that indicates that these T cells in the tumor, particularly with the combined knockout will be more sensitive to levels of cytokines that are in the tumor microenvironment. It's -- they are actually very, very large global changes as well also as with chemokines. And I think this also makes the point that, we hear a lot about tumor immune exclusion and the idea is if you could only get T cells into the tumor, then you could potentially enhance antitumor activity. One of the best ways to do that is to increase the level of the chemokines, which are essentially the chemoattractant cytokines that are being made in the tumor, which again, some of which go up, in particular, some of the -- I won't bore you with the particular names because there are a lot of them. But these are not individual cytokines. They're global changes that represent banks of cytokines that are commonly regulated by transcriptional programs probably more than you wanted to hear.

The next question comes from Jeff La Rosa at SVB.