Good afternoon, everyone. This is Daina Graybosch. I'm a senior equity research analyst here at Leerink Partners, and I'm excited to be hosting, I think for the first time at this, the management team from Xilio, to talk about their novel conditional immune, immuno-oncology activating, pipeline, Rene and Uli. I think that this morning you had some new disclosures, so gonna throw my question guide out the window mostly, and we'll do it a bit ad hoc, but that's how I like to do things anyway.
More fun.
That's fine. It is more fun. Yeah, why don't we start with that? Why don't you talk about what you newly disclosed today, and then we'll get into it.
Excellent. Thank you. Thank you for having us. Yeah. We have been in this space of masking technology for a while, and very excited to see sort of masking really come into sort of appreciation as we learn where masking has made the most benefit, and most recently, we're seeing that with T-cell engagers and being able to give very potent T-cell engagers that are masked. Sorry, hang on. Oh, no. Okay. Yeah. Thank you. Okay. Seeing masking really make an impact in the clinic and cell engagers has been very exciting. We have been positioning now our masking technology, which is clinically validated in about 300 patients, multiple molecules, into these more complex multispecifics and cell engagers in particular.
As we've been watching the competitive landscape really evolve around masked PSMA, we had two programs in development, one for masked PSMA, one for STEAP1. Really kind of rising to the competitive landscape, we just this morning announced at our prostate program, we're combining the two. We're losing no time in the timelines, and we are moving into a multi-targeted PSMA STEAP1 cell engager that has masking on the CD3 and has co-stimulatory signaling built into the molecule. With all of that, right, we think we can address things like durability of response, antigen escape, and maintain that tolerability through masking. Very excited about this. Big kudos to Uli and his team. Not easy to pull off a molecule like this, and again, not losing any time in terms of our IND plans. Do you wanna add to that?
Maybe just adding briefly to this. For us, it was always a main objective to have a modular platform. That is to have components that could be really optimized for the particular product in question, taking into consideration what are the target properties like expression, normal tissue burden, to then make a decision on how you design that molecule. Now we feel that we have matured each one of these components where it is truly a modular platform. In the case of the prostate program, that modularity allowed us to progress in parallel, PSMA and STEAP1 and a dual targeted molecule, and then come to a point of decision where we felt dual targeted is ultimately the best molecule to advance.
I have so many questions. let me start with some basics. masked on the CD3-
Yes.
Is either STEAP1 or PSMA masked or the costim?
For this molecule, only the CD3 is masked. The TAs are not masked, the costim is not masked. This is based on the data as we brought this molecule through non-human primate testing, that this had the best profile.
Can you tell us what the costim is? Does that have activity then as a tumor-targeted costim even when the CD3 is not unmasked?
Mm.
Yeah. We thought long and hard around what co-stimulatory signaling to incorporate into a T-cell engager. We knew we wanted it to be tumor selective and dependent on TCR signaling.
Mm.
We wanted to ensure that this is a co-stimulatory pathway that is maintained in aged T-cells. We made the decision to advance the PSMA, STEAP1 program with CD2 co-stimulation. Now, what is CD2? CD2 is a very important co-stimulatory pathway. It's also an adhesion protein that's involved in organizing the immune synapse. Importantly, its signaling is really dependent on TCR stimulation. The affinity of that ligand is also designed such that it does not signal in the absence of unmasking of the CD3 and clustering.
Wait. Okay, say it again. It's in the immune synapse. Say that last bit again. I was interpreting the first part, and then I need to hear it again.
Yeah. It really doesn't by itself signal unless it's in the context of TCR stimulation. Because we have a very efficient mask on our CD3.
Mm.
We don't get ectopic T-cell activation in the absence of the CD3 unmasking and binding to target antigen. That's really corroborated now by primates tolerability data. We were able to dose our dual targeted PSMA STEAP1 TCE with co-stimulation built in at 3 mg per kilogram in the primate with no clinical signs, no CRS. That's what's giving us confidence that indeed the molecule's activity is dependent on unmasking in the tumor microenvironment.
Two things, both on the safety and the efficacy side, if that T-cell's receiving TCR stimulation through some other route, let's say another antigen on the tumor cell that's not STEAP1 or PSMA, can the CD2 signal give you an efficacy potential or are you really not seeing that without the really strong CD3 signaling?
We haven't looked into that, but of course you could imagine that as part of antigen spreading this could only be a plus, right? The ligand, the natural ligand for CD2 is CD58 and the natural ligand is actually commonly lost and deleted in certain types of tumor as a mode of immune evasion. We know it's an important pathway, but we're with our co-stimulation, we're only adding back in the ligand that's often lost. This is not a super potent ectopic type of co-stimulation like CD28 that has been marred in the past with toxicity. CD2 is in a very different class of co-stimulatory domains and that's in part why we chose it.
Got it. CD2 has been used in other clinical programs. I know Fate knocks out, or uses CD58 as the allo evasion.
Mm.
That's the opposite. Do remind me, I'm pretty sure somebody has used CD2.
Yeah.
Maybe in a CAR.
Novartis put forward.
Yeah.
T-cell engager target for-
That's what it was.
...for some hematologic malignancies in the past.
That's in the clinic, right?
I believe so.
A trispecific?
I believe so.
Yeah. It's like their CD19. Okay. See, this was on the spot, so I was trying to remember. They might, you might get some clinical validation of this approach from them in a different context prior to yours coming in.
Yeah, I think here again the important part is there are obviously always going to be differences between the molecules. We really try to ensure in this molecule that each one of the domains is optimized and so that goes for the CD3 domain as well in our PSMA STEAP1 T-cell engager. Because we have effective co-stimulation, we actually looked at various different affinities of the CD3 as well and took that into consideration when selecting the right final configuration to move forward.
What is that? It's a reduced affinity or a different epitope?
Yeah. I would call it medium affinity and we have learned when it comes to masking, this is overarchingly their really important balance between, on the one hand, having a very effective mask and on the other end ensuring that you have the right level of potency because I think what we've seen in the field is there are different masking solutions and also folks come with their baseline molecule at different potencies and there are various different ways on how to achieve ultimately I think a clinical profile that's desirable. Our approach has been to create a very effective mask and then optimize for the potency and that's I think true across the portfolio.
Mm-hmm. I agree. Can you talk about the mask and this is... I think you are doing it fit for purpose.
Mm-hmm.
What is unique or differentiating about the mask and the linker that you're using in this case from the two others that are in prostate let's say or from your other programs?
Maybe I'll explain it sort of from the clinical perspective and then Uli can get into the scientific piece. I think when we think about the other two programs right in the PSMA space and the masking they're quite different and they both can work right-
Yeah.
... when we think about masking with our prostate molecule in particular 'cause each one is different I'd put us a little bit in between those two other molecules right? You have one that is using steric masking right which is maybe a less effective or an early gen mask meaning that you can still get binding when it's masked right at some level, but they also have as Uli mentioned the detuned CD3 and detuned PSMA so all of that balance works and then you have on the other maybe extreme end here is full potency CD3 and PSMA with a mask that interacts with the molecule.
Mm-hmm.
Our mask I would say is a little bit more effective in terms of the therapeutic index it can widen there in head-to-head in vitro studies and our CD3 is detuned a bit so I feel like that puts us kind of in between the two if that makes sense and that's specifically for our prostate program. Each program is sort of a little bit different.
You bind a little bit unmasked?
I would say the mask is more effective, so less binding when masked but readily cleaved, right, what we call efficiency of masking, so when it's masked you get no binding or minimal binding, at some level you would always get binding, but better, I would say, than any of the current ones in terms of the efficiency of the mask, and then when it comes off it comes off but a slightly detuned CD3 as well, so that's really what's giving us the ability to add STEAP1, avoid CRS right in the primates, and add the co-stim. Does that make sense?
Yeah.
In between.
Well tell me if this is right.
Yeah.
I think the beauty of the ones that bind a little bit and there's others like one of the CTLA-4's is the same way is that you can build a local concentration-
Yes.
... and maybe even though you have a lot of mask you can still build that what's needed to really get a productive T-cell you're coating the tumor and so.
You need the unmasking to happen very efficiently and quickly, yes.
Yeah.
Yes.
Yeah. You can build it up you sort of almost have a depot of masked things in your tumor and then even if it's a little bit less efficient maybe it takes longer to ramp up but you've sort of built coating to then bring the T-cells in.
We do.
You're doing that with your two antigens.
Mm-hmm.
Does that help? Versus let's say going after just STEAP1 or PSMA, right? How does that change the concentration or the distribution of your molecule?
Mm-hmm. I think Uli will answer that, but before that, I would just remind you the costim itself, right? Really, I think is important here because with costim, you can have efficacy at lower concentrations as well of both.
Yeah. To your question on STEAP1 and PSMA, one of the key reasons why we wanted to pursue a dual targeted molecule is that actually 30% of prostate cancer patients at advanced stage disease are PSMA negative.
Mm.
More than that are quite heterogeneous in their expression of PSMA. In a PSMA negative patient, the chance of seeing a response with a PSMA targeted agent is not great. Our goal was how do we widen the potential patient population that could benefit while at the same time cutting off routes to resistance.
Mm.
The routes to resistance through antigen escape or outgrowth of negative clones, that's significant. It's sort of a general theme in cancer biology. We think it's limiting the durability of current prostate targeted therapies. The dual targeting therefore is to drive more durability. Costim on top of that is the second factor that we think can further enhance that.
Are you driving more selectivity?
We designed the molecule to be an OR gate, not an AND gate. What that means is that the molecule is able to bind either STEAP1 positive T-cell, tumor cells or PSMA positive tumor cells. It does not necessitate that cell to express both. One could design a molecule like that, but then you would not be addressing the negative or single target expressing population. Where that could matter though, is that there are by having essentially two binding arms, you have a bivalent molecule, right? There's therefore an avid interaction. If there's a high level of target expression of both of these targets on tumor, which in a certain population is the case, then you would bind that probably preferential relative to some tissues where only, for instance, STEAP1 is expressed.
Yeah. Yes, it's an OR gate, but probably with some selectivity based on the avidity.
Our primary goal was, again, to drive towards durability and minimize resistance, but the added selectivity, we think it's a potential added benefit.
At least you're slowing the off rate, which then gives time for your mask to be cleaved. Is that fair?
Yeah, I think that's fair.
Yes.
I mean, there are other ones you could try here. You could have tried KLK2 or DLL3 if you wanted to go for sort of the neuroendocrine conversion.
Yeah.
Why these two in the end?
Both of these have clearly shown some significant anti-tumor activity, right, when used as single targeted, TCEs. We do think that the majority of the population is what we want to hit, and we felt that by doing a, you know, bioinformatic analysis between STEAP1 and PSMA, you're hitting 95%.
Mm.
The neuroendocrine, it's an important subtype, but in prostate it is only about 5% to maybe 10%, right? That's where we focused first. You're absolutely right that the concept we do think is.
We have feedback. Go ahead, turn your.
Yeah. Is this back on? Great.
Yeah.
I mean, as I was saying, ultimately we think that the concept of dual targeting is something that is applicable to not just STEAP1 and PSMA. I think there are a number of tumor associated antigens that are expressed in similar lineages, similar tumor types, where again, you might not have 100% high homogenous expression. The approach with dual targeting, addressing that outgrowth of antigen negative cells, that could be there, very relevant as well.
I think what J&J has at least early demonstrated in multiple myeloma by making the trispecific GPRC5D, BCMA. There you don't have necessarily the case. You're getting good expression on both of those. Is somehow having that dual antigen targeting improves the efficacy, durability, and reduces the toxicity? I wonder if is that gonna be unique to myeloma, or is that something we'll see with all dual targeting, and have you already seen that in your preclinical, and do you have a hypothesis on what's happening?
I think what I can say, and Uli will jump in, is because we have this modular platform, we are bringing multiple constructs of these molecules all the way through NHP, and this molecule and this construct had the best profile. Maybe there is something to that.
I look, I think the hematological malignancies have been at the forefront of T-cell engagers. That's where we initially learned that the approach can work. There are some important differences in the disease biology. I would not yet lean forward as much to say that this is going to be 100% translatable. My confidence is more from a cancer evolution standpoint, that if there is loss of antigen expression, if there are heterogeneous expression patterns, if you get at that with a multi-targeted molecule, your chances that you're gonna prevent relapse are gonna be bigger. We're gonna have to see how this pans out. We're currently working towards preclinical models that reflect some of the heterogeneity, so we can answer those questions maybe at least a little ahead of getting to the clinic.
I always worry about when we get to these higher order specific, that it's just very hard to balance avidity, get the right affinities of everything, get the right geometry, and not have ADAs.
Mm-hmm.
I wonder, like, how much time did you spend optimizing it? Is my tourism just a little bit too conservative, because you're going with this tetraspecific?
Yeah. Our approach in the very beginning, right, is really focused on clinically validated components, manufacturability, low immunogenicity, and I think we've now shown that across several molecules. Again, you know, we would tend to think, right, a more complex molecule may come with more challenges, but it turns out this construct, this 3D confirmation, this molecule performs really well. I think that's the key, is that we didn't just go with it. It came out of the process, looking at all of these things, including immunogenicity and manufacturability, and it was the clear winner, and it had the profile that we hoped for.
Yeah, that's right. The approach that we take, it's about 3-5 design cycles, and you can think of each design cycle being over 100 molecules. You design this drug discovery funnel to look at architectures and geometries, right, initially with sort of representative domains. We really get down into the specifics of, okay, what's the right property of that domain at that location? A lot goes into it. We've miniaturized and parallelized as much as possible. We've been actually quite efficient at progressing.
Because you've really sampled a wide design space, you then have actually, I think, ultimately more confidence that when you get to development, when you get to CMC, and you've looked at pre-developability, accelerated stability, you've done a lot of the de-risking activities that you have a molecule that can go all the way.
I think to believe this, you have to really believe in your non-human primate model and its translatability. Why should we believe that the one that rose to the top there is gonna be the best one in the clinic?
Yeah. We are looking obviously for importantly, signs of healthy tissue activation, right, of the mask, CD3 binding, CRS, and that's what we're really looking for in the NHP. We're also relying on, right, other data that's out in the public domain. Based on what we've seen to date, we're very happy with the tolerability profile in NHP and the dose we've been able to get to-
Mm-hmm.
...the therapeutic index window and where we think that will lead us in terms of the clinic.
Yeah, I would say that, you know, the utility of preclinical models is very much tied to what's the question that you're asking of them, right? If this were a completely novel antigen that nobody has yet taken to clinic, there are certain, I think, questions around early preclinical data including tox data. This is not the case, right? This is a clinically validated target with a clinically validated modality, and we've seen really encouraging tolerability. Then you pair it with the efficacy side, where we have seen activity sub-mg per kg, so really at exposures that fall significantly below the exposures that are tolerable in primate, and then you're not just looking at dose. You're really looking at exposure window, and that's what we've seen, positive TI for both our prostate program and for our claudin program.
I think that's underpinning that the masking solution here seems to be working.
With the claudin program, are you gonna also add costim, given costim engagers alone have activity with claudin?
Yeah. I think the way we approach all of our cell engager programs and with this modular platform that Uli and his team have built is we bring multiple constructs along in parallel. Our claudin 18.2 could be a simple masked TCE that we call ATACR. We are bringing a molecule along that has costim built in. We are bringing a molecule along that is single masked and dual masked. Again, just like we did, one of these will make it all the way through into the IND-enabling work very soon. We have that opportunity. We really let the data drive that decision.
When are these both gonna be in the clinic, and when do we start to get to that clinical de-risking data? Because it's validated, sort of can you get to that faster?
We are planning for INDs for both of these cell engager molecules in 2027. We do believe with the NHP data we've seen and our experience in the past with masking is we should be able to start at higher doses in the clinic than you can without a mask. That should be able to help us accelerate clinical development a bit. I think, you know, that plan will be obviously what we see in the GLP talks and with the FDA on deciding what is that starting dose. That's been our experience with masking, is they really do understand it, and that has allowed us to start at higher doses in the clinic.
Data in 2028.
INDs in 2027 and data to come as soon as possible.
Got it. Okay. Maybe let's spend the last three minutes on your CTLA-4 antagonist, which you have, you know, interesting data, compelling data in microsatellite stable CRC, as well as a potential biomarker in plasma tumor mutational burden. I think you are currently waiting to find a partner for that asset to bring it forward. Can you give us an update on that partnership and what's anything that's gating that partnership from your side in terms of like milestones or a type of partnership you'd be willing to accept?
Yeah. Our CTLA-4, this was the first program we put into the clinic. It really validated the masking technology. We've demonstrated now in over 100 patients that we get about 70% of this molecule activated in patient tumors, up to 90% consistently, less than 15% active in peripheral blood. That has really been helpful in validation and has led to some of the other partnerships that we have as a result of this molecule. We have demonstrated now in late line MSS CRC a 26% overall response rate in patients without liver mets. Most, I think, exciting for the field is recently we identified retrospectively a biomarker that really wasn't sort of in the limelight, if you will, for MSS CRC, and that is plasma TMB or tumor mutational burden.
In the past, as a field, we looked at tissue TMB. It did not predict outcome for immunotherapy in CRC. It turns out that the plasma-based TMB, for a lot of reasons, was predictive of response with vilastobart and PDL-1. We were able to get that response rate in that subgroup up to 40%, which is really meaningful. We have, I think, a very differentiated tolerability profile with the masking consistent with that 15% or less in peripheral blood. Much less severity of side effects, less rate of side effects than unmasked comparators. This is, I think, prime for combination therapy. We think it makes a lot of sense to combine this with PD-1 bispecifics in CRC. Other tumor types, like non-small cell lung, is another area where we've seen activity with this molecule.
We like the idea of combining it with our own PD-1 IL-2, potentially our own T-cell engagers as well. I would say there's nothing gating a partnership per se. I think there is renewed interest in this methodology based on the data we showed in the high plasma TMB population. We are pursuing actively a partnership because really this should be expanded into a broad range of tumor types that really does need a big pharma partner to fully pursue. That's underway, still, having those active discussions. We also like this molecule very much as a combo agent for many of the things in our pipeline if we can keep that and carve that out. Do you wanna add anything?
Yeah.
Can you give an update on the partnership with Gilead for the IL-12 and any milestones that could be achieved this year for that program?
The remaining milestone for the IL-12 program is the opt-in on the molecule, which is associated with a $75 million opt-in fee. We expect to deliver that phase I, II data package to Gilead in the first half of next year. They have some period of time, right, to evaluate that data package and then opt in on the molecule.
The PD-1 IL-2 program. Can you talk about where that is?
We're very excited about the PD-1 IL-2 program. This is PD-1 IL-2 with a masked IL-2. If you're following the field, there is an exciting molecule by a company called Innovent that is a PD-1 unmasked IL-2. We were really excited about the deal they just did with Takeda, which is I think a $1.2 billion upfront with $10 billion in milestones with pretty early data for that molecule. I think what was exciting is they showed 30%-40% response rate at their higher doses in a post PD-1 setting, in tumor types like lung, melanoma, CRC. What we're able to do with our masking technology is bring a full potency IL-2 with a mask, which we think can improve the efficacy and also maintain or improve tolerability with masking.
You wanna add anything, uLI, to that?
That's in the clinic this year?
IND midyear. That's right.
Yeah. This is going to be the really the next program in the clinic. As Rene pointed out, there are some key differences relative to Innovent's molecule, namely the potency. Ours is bivalent PD-1 as opposed to a monovalent PD-1. More importantly, our IL-2, much more potent, didn't have to detune it because we have the mask. Preclinically, we've demonstrated, it looks like it's gonna have best-in-class potential.
It's a highly efficient mask, similar.
Yes. Similar masking strategy.
You can avoid 'cause I think, like, Roche's, you know, bispecific IL-2s, you always ended up with just so much being driven by IL-2 binding, even with a reduced potency one. It was just sitting around in the spleen. By masking, you can really drive that concentration into the tumor.
Absolutely. We've seen that. We've made a molecule that is the same but doesn't have the mask, and you see rapid clearance, you see IL-2 toxicity in the periphery. With our masking solution on it looks like a PD-1 antibody, right? Long half-life, no IL-2 tox. In the tumor, there you get expansion of antigen-specific cells, expansion of the progenitor stem-like T-cell pool that's really important for durability. It's clearly having the effect in a more focused way in the tumor microenvironment.
That's great. Well, thank you. That's exciting. A lot coming. Thanks for the questions from the audience. Yeah.