All right, we're going to go ahead and get started. I'm Stephen Willey, one of the senior biotech analysts here at Stifel, and glad to have with us for the next discussion the CEO of Pyxis Oncology, Lara Sullivan. Lara, thanks for coming.
Thank you for having me.
Maybe we can start with a quick overview of the company, kind of who you are, what you do, and then we'll jump right into some Q&A.
Pyxis Oncology is an ADC-focused company. We have a lead asset in clinical development, which is a novel target addressing a target called extracellular domain B, or EDB. It's a splice variant of fibronectin, and it's a noncellular target in the tumor microenvironment. Our ADC is an extracellular cleaving ADC. The ADC addresses the target in the tumor microenvironment, cleaves in that extracellular matrix, and then begins to kill tumors through a three-prong mechanism: direct tumor killing, bystander effect, and immunogenic cell death. We brought this asset into the clinic in early 2023 and read out our dose escalation data of a basket trial last fall, last November. We showed tumor regression in six tumor types with a clear, strong signal of a confirmed ORR of 50% in head and neck. We started our expansion trials at the 5.4 mg/kg dose earlier this year in monotherapy.
We also started a combo therapy trial in combo with Pembro.
All right, very good. You mentioned MICVO, though. This is kind of the centerpiece of the story here. Obviously, the first in concept approach to ADC design and payload delivery. Maybe you could just talk a little bit about it. I think you referenced this a little bit in your intro comment, but what kind of specifically differentiates this type of ADC from a design and mechanistic perspective? What do you have from a translational data generation perspective that supports that?
Yeah, so the clinical ADCs that are in development and have been approved have taken the canonical approach that's been in practice over the last really several decades of being cell surface targeting ADCs, meaning the assumption had been that for an ADC to work, you have to address a target that's on the cell surface and that the ADC addressing the target, basically the ADC then gets internalized by the cell and cleaved by the cell inside the cell, inside the tumor cell. That's been conventional wisdom. To your point about our first in concept approach, our target is in the extracellular environment. It's in the tumor microenvironment, and it cleaves out there in that tumor microenvironment, in that matrix.
What's different about it is the location of cleavage and the fact that the payload is freed up in the tumor microenvironment and then kills the cells directly coming out of the matrix. It sounds very simple and straightforward, and it is, but it's different. One of the things that we have shown through our recent translational biology work is the presence of the proteases that cleave the ADC. We've proven them out to be in the tumor microenvironment. We've had 10 biology and translational posters put out over the last four or five months, six at Triple Meeting, two at ESMO, two at AACR, all of which point to proof of mechanism for this novel ADC.
I would encourage people to check those out because a lot of the conventional wisdom that you need to internalize your ADC to directly kill the tumor cells is just not proving to be the case, that in fact, bystander effect, immunogenic cell death, these things that happen in the matrix are a core component of how our ADC works. This translational work supports that.
You talked about bystander effect, and I think for a lot of investors, ENHERTU was kind of this gateway of understanding what kind of the real clinical and commercial value of bystander activity is. How would you characterize the type and the quality of the bystander effect that you're seeing with MICVO relative to something like ENHERTU?
Yeah, so I think ENHERTU has delivered tremendous impact and taught us all a lot about ADCs. One of the most recent insights I think that's been gleaned from ENHERTU is it was done by some researchers at Duke. I think it was published in April of this year where they showed why ENHERTU works in low HER2 expressing patient populations. If you're a cell surface target, again, the conventional wisdom has been you have to have that target expressed to a certain degree for the ADC to work. In low HER2 expressors, the ADC was working, and people were wondering why. That's because of the bystander effect, where once the payload is cleaved and it sort of triggers almost like a pinball, that's how I think of it, like a pinball in a pinball machine.
It's just bouncing around and triggering more and more killing of cells. That actually is a key component of ADC mechanism. In HER2 showed this with the low HER2 expressors, which we see as a proof of mechanism to our ADC. I think they're very comparable to each other because of that ramped up or amped up component of bystander effect contributing to the overall tumor regression.
Okay. So we saw the first- in- human clinical data for this drug, I think it was about a year ago now.
Correct.
What would you kind of highlight, I guess, as the key learnings from that preliminary data disclosure? What elements of the data you presented did you think were the most interesting and the most validating?
We were pretty excited to put that data set out last year, last November. We had dosed 80 patients, and we had data available for 77 of them. It was, as I mentioned earlier, a basket trial, dose escalation. We had nine tumor types recruited across dosed across 10 or 11 dose levels. We were learning a lot of things simultaneously around what the efficacious dose is, as well as which tumor types were demonstrating the strongest responses. When all was said and done, we found an efficacious dose range of 3.6-5.4 mg/kg, which is a reasonably high dose for an ADC. That was very well tolerated. We had a modest set of AE profiles related to the kind of payload component of safety that you monitor, which includes things like neutropenia, neuropathy, ocular tox, cutaneous lesions.
We had no Grade 3s treatment related for ocular toxicity or neuropathy, and we just had a modest amount of cutaneous and neutropenia. The safety profile showed well compared to the other ADCs that are out there. From an efficacy perspective, we showed strong tumor regression in head and neck with a confirmed 50% ORR. We had a confirmed CR and two confirmed PRs. This was in both HPV positive and HPV negative patients. We did not see a differential response based on HPV status. I think that's really important given a lot of the evolution that's happening in the head and neck marketplace. We also saw tumor regression in breast, sarcoma, ovarian, and lung as well.
I think that broad tumor regression supports the key point around an extracellular targeting ADC because the extracellular matrix exists in every solid tumor type that's out there. The question is, in which solid tumors could this particular mechanism be effective? We saw that happening in multiple tumors, and we saw it very pronounced in head and neck. Our translational work recently has actually identified that one of the reasons it was so pronounced in head and neck has to do with the way the noncellular elements are organized in the tumor microenvironment. We put out a poster recently that showed in the case of head and neck, the fibronectin fibers in the tumor matrix are very well organized, almost stacked in parallel, and the payload can traverse right through the tumor cell very easily with no barriers.
In the case of pancreatic, where we showed less tumor regression, those fibers were tangled and chaotic, and so the payload couldn't traverse as easily. We have in some that study really showed an agent that has broad potential where that potential can be particularly powerful in certain tumor types like head and neck, where the characteristics of the tumor line themselves up nicely for an ADC that is cleaved in the extracellular environment.
Okay, interesting. I guess of the three head and neck cancer patients that you saw achieve a response, you had a couple of interesting anecdotes, right, where you had a patient who had actually achieved a complete response, came off drug for a little while, that response was preserved. You also had a patient, I think, who had a fairly deep PR, came off drug for a little while, and then that response appeared to further deepen while off drug. What do you think is kind of biologically happening here with this?
Yeah, so one of the insights we got from this 80-patient study was we looked at prior treatment patterns for our patients, and we saw that this agent worked in patients regardless of prior therapy. We had responses in patients with prior eGFRs, with prior taxanes, with prior other ADCs, with prior IO. To your point, in the actual study population itself, we found that patients who were treated with our agent and then paused and restarted still showed responsiveness. There is something, I think, about the tumor matrix, maybe the stromal architecture itself, the maturity cycles of the stroma, that seem to allow it to retain susceptibility to this agent regardless of what the patient had been treated with before.
In the case of the patient who sort of paused and resumed and had a deeper PR, essentially that patient had a nice PR on 3.6 because we were early in the escalation. When they paused for a few months and then we had cleared ultimately the 4.4 dose, they came back on at 4.4, and their PR deepened from like -50% to -90%. They almost got a CR. I thought of that 4.4 as the Goldilocks dose for that patient. In the case of the patient with the CR, they had personal reasons. They were relocating. They were moving to a different continent. Logistically, it was hard for them to stay connected to the study, and so they ended up ending therapy.
We saw them six months later when they moved back to the U.S., and their CR had still held six months without therapy. These anecdotes are the things we rely upon in early stage development where we do not yet have ends of hundreds of patients, but I think are very powerful concepts that help us as we continue to develop the agent going forward.
You talked about safety tolerability, and I think this is an auristatin payload, right? I think there is kind of a very consistent characteristic AE profile that we have seen historically with auristatin-derived payloads. This does not appear to look like that in terms of no real heme tox, no real ocular tox. Was that intentionally designed into the molecule as well? Did you know that this was going to have kind of as wide of a TI as it appears to have?
No, we didn't really know what we would ultimately see in terms of the TI or even the dose level until we started experiencing it. What we did know was the same ADC construct, the same linker, the same payload, the same conjugation chemistry. Pfizer had built this ADC. We licensed it from them at development candidate stage. We had seen that same ADC construct Pfizer had used in their HER2 asset. In the case of the HER2 target, which is a different target, different mechanism, Pfizer was not able to dose comfortably above 4.0. We thought our dosing range would look more like Pfizer's with HER2, and instead we were able to dose up to 5.4. We also didn't know which of those features of AEs would be more predominant in our agent or less predominant.
What's interesting is in our November materials last year, we took a look at our 5.4 dose level compared to the approved ADCs that are out there and their AE profile on the label. Even with common payloads like MMAE payloads, each ADC sort of has its own thing. There is something that's a little more complex than probably just the payload and the target. The interaction between the payload and the target is more complex, I think, in determining the AE profile than maybe we might have initially thought. The Pfizer scientists did a brilliant job designing a beautiful ADC. They optimized it for permeability, for potency. We have a very stable DAR of 4. It's very homogenous, and we have really great site-specific conjugation chemistry. This ADC holds together.
It does not break apart when it is traversing, which is something you see with Padcev, for example, and that can cause havoc because the payload can go all over the body. All those features, I think, really lend themselves to a nice ADC. Ultimately, to your point, which AE profile kind of emerges, you start to learn that more as you get close to that MTD. Because by definition, if your development strategy is to dose up to the MTD, you are going to have some AEs because the AE level is what tells you to stop. Which ones they are for a new target is interesting to see emerge. We were fortunate because I think the ones we experienced in part one, like the neutropenia and the rash, were treatable and reversible and manageable. Those are things you can do something about.
Okay. Maybe you could talk a little bit about the rationale that underlines the design of the cohort expansion effort that you have ongoing now. I know you're guiding to having some preliminary data disclosure from this. We'll talk about that in a minute before the end of this year. Why did you choose kind of each of these cohorts, either for monotherapy and/or combination-based development?
Yeah. So we were very pleased with that head and neck signal being as strong as it was in part one. We still see a lot of unmet need in the head and neck space, even with the recent competitive activity there. Head and neck has been underserved for 20 years. In the last two years, between Merus, Bicara, now Johnson & Johnson, now Corbus entering, patients are getting some options. These are not patients that are cured, and these are not 100% response rate agents. There are still patients who will not be served by these current agents or patients who will relapse on these current agents. We looked at that and said, the market right now is primarily the current standard of care is primarily driven by platinum and PD-1.
Tomorrow's market may be driven by a new standard of care that's eGFR-based with Merus and Bicara . Best for us to design a trial, a monotherapy trial, where we can segregate our patients into two arms based on their prior treatment therapy. We can generate data that is comparable to today's standard of care and then data that can live and show what place we have in tomorrow's standard of care. Arm one data that we are generating is testing our agent in the platinum and PD-1 resistant patients. These are today's standard of care patients. That data set will be directly comparable apples to apples with the Merus, Bicara , Corbus, J&J data sets that are out there. Arm two is the eGFR and PD-1 resistant population, which is where Merus, Bicara , J&J are going, and today's Cetuximab patients.
If one believes that those eGFRs become the early line choice of therapy, then this data set we're generating in arm two will show how our agent works in patients who have been previously treated with eGFR. It provides a market entry strategy right now that is essentially unserved because there is no line today that comes behind the eGFRs. If the eGFRs come in first, there's a wide open space. Our data set from that arm two will address that. That's our monotherapy program, those two data points. Our combo therapy program is our MICVO agent in combo with PD-1 and for first and second line patients. This data set is also going after that same population that Merus and Bicara have with their combo data that's out there.
We think that with our compelling monotherapy data that we showed in part one and our mechanism, which relies on immunogenic cell death, which is recruiting the immune system, that throwing Pembro on there in combo could be very compelling for these patients. We know that ADC plus IO combos are very powerful. We saw that in bladder cancer with Padcev plus Keytruda. That is the rationale for us going into the combo fight as well.
Okay. As it pertains to this year-end data disclosure, the update, what is the expectation framework that you want to establish here going into this disclosure just with respect to the scope and the level of interpretability that we should expect from the data that you intend to present?
Yeah. I think we were pleased to see both the J&J and Corbus data presented at ESMO in addition to the Merus and Bicara data that's out there because it demonstrates a market that's really aggregating around some effectiveness levels that any entrant, Pyxis or anybody else, should pay attention to. In the case of monotherapy, I think what's really been established by these players is a 35%-45% ORR is what you need to put up to be competitive with them. If you're going after that platinum and PD-1 resistant population to be a credible competitor from an efficacy perspective, you need a 35%-45% ORR. Anybody needs that. If you put that up, then the question becomes, okay, how do you compare with these competitors from a durability, from a safety perspective?
That is where early data sets need to begin to mature to be comparable with one another, right? We do not know the answer to all of that yet because all those studies are in progress. In the case of combination therapy, the Merus and Bicara have put up roughly 60%-65% ORRs. Beautiful ORRs for that patient population. I think, again, the minimum of what you need to see to be a credible competitor in that space for that frontline opportunity. In the case of our monotherapy arm that is targeting the eGFR-resistant population, remember, this is where we think standard of care is going. It is not there yet where we think it is going. Because there is nothing right now available to come behind the eGFRs, there is no established expectation yet for that line.
In fact, we haven't really heard a lot of dialogue about it with respect to the marketplace. We hear dialogue about it from PIs who say, "Oh gosh, there's nothing there. Maybe we'll throw chemo." Whatever we have available to us for patients at that point is probably a modest 10%-15% ORR. If somebody can bring something 20%-ish to us to be post-eGFR, we'd use it. I think that's we sort of see the 30s in the platinum and PD-1 resistant population as the ticket to play. We see the 60s in the combo with PD-1 as the ticket to play. If you rely on the investigators, 20-20s is what you'd need to put something behind eGFR to move the field forward.
Your combo cohort is enrolling both front and second line patients.
First and second line.
Right. So I guess in the context of also having some second line patients into that cohort, then maybe that 60%-65%, which has been generated in a pure frontline population, is a little bit of an aspirational target.
Yes. It could be a little more nuanced when we actually break it out by line to your point. Yeah.
Your point is well taken on the post-eGFR second line monotherapy, right? Because we do have these single agent benchmarks in second line, but the competitive relevancy of those benchmarks within the next two to three years is probably not meaningful.
Right. That's right.
Because those drugs will find their way to frontline.
That's right.
How far are each of these cohorts away from their full target enrollment? I think you're targeting 20 patients per cohort.
Right. We are targeting to enroll exactly 20 patients in each of the arms. That would lead us to a total of 40 patients in monotherapy enrolled. It is harder to give enrollment targets on a combo in escalation because you never know exactly how many patients you will have at each level. We would like to enroll roughly 20 patients of head and neck in the dose escalation for combo because combo is multi-tumor at the escalation phase and backfill focused for head and neck. Those are the enrollment targets. In terms of the data disclosure, we have not guided on how many patients or data points we will put out for each of these three pools of data. We are indicating these are preliminary data cuts. They are in different levels of maturity because they all have had different enrollment curves at different levels of maturity.
They're each running on their own pace. The way we're thinking about disclosure is that regardless of the end that ultimately ends up in each of those three pools, we will provide a data set that we believe is interpretable within each of those pools. Also, when you look at those pools collectively, the start of the gestalt, so to speak, of seeing how this agent performs in three different settings should be informative.
Okay. Will there be any biomarker or translational data in conjunction with this, or is that probably a later time point?
Yeah. I think ultimately, the way we handled the translational data was primarily through the posters this year that we've put out on the proof of mechanism and sort of the why head and neck story. We do as much translational work as we feasibly can with paired biopsies for patients. We can't always get them anatomically for head and neck patients. Sometimes they're not accessible. You really build that patient-driven translational data set as you go. In terms of a biomarker, we think that ultimately it is unlikely there'll be a single biomarker that will encapsulate predicting a response for a MICVO patient because ultimately we think the response is dictated by multiple factors. It's hard to aggregate that into any one particular biomarker or even series of biomarkers.
For example, the things that we've seen correlate with responsiveness include target expression, protease concentration in the tumor matrix, acid-base, pH level, pressure in the tumor, that pressure gradient, the orientation of the fibronectin fibers. Are they open that you can traverse like this, or are they tangled? It's hard to imagine how we kind of create a composite biomarker on that. What we can do is look at those features and then use those insights over time to look at other tumor types as well and say, "What other tumor types histologically or pathologically look like head and neck?" That could point to the next horizon of value.
It would also seem to me that your ability to kind of interrogate that combo cohort with Keytruda would also maybe inform some of these other non-head and neck cancer opportunities as well.
Yeah. That's a great point. That's a great point because we still have those other tumor types in the combo escalation. It is another way to do signal finding clinically. We did put some other tumor types in combo escalation that we did not test in mono, like gastric, for example. There are a few tumor types in there that we're learning as we speak.
It's probably a better question to ask after we've seen data. What could next steps look like? I guess presumably there's still some dose optimization work that you need to do just to kind of check the Project Optimus box. If I look at Merus, for instance, and kind of how quickly they were able to go from, say, like a phase 1B cohort expansion in second line to a registrational trial, is that a fair timeline, do you think?
Yeah. No. I mean, Merus really set the standard for us all, right, in terms of execution and speed to data and insights and turning it into an actionable development path. With the data disclosure, we will include, in addition to those three pools of data, insights into the market opportunity for each of those segments that these three pools can point to because ultimately we think that the head and neck marketplace is more nuanced than maybe the dialogues have been to date. I think that's really just reflecting a market that's beginning to mature and evolve, right? In the early days, it was Merus versus Bicara. That was what the dialogue was. There was not anything else there. Now there's Corbus. Now there's us. Now there's Johnson & Johnson. Now there's two mechanisms, eGFR and the ADCs. There's different lines. There's different HPV status.
There's maybe different correlations based on anatomical location, etc. Ultimately in the data disclosure, we will sort of pull those factors together with the data, and then we'll comment around the development path forward, including regulatory engagement strategy and early insights into timelines associated with that plan. Stay tuned with the data for more insights on that.
Last question. Just cash runway, what that allows you to execute on.
Yeah. I think we just put out our recent Q, and it's $77 million cash balance with runway through or into second half of 2026 through the catalysts that we're putting out. Yep, we're excited to get the data out there and continue to advance the company.
All right. Lara, thank you very much. Looking forward to the year-end update.
Thank you so much.
I'm sure we'll be touching base.
We appreciate it. Thank you.
Thank you. [audio disruption]
This presentation has now finished. Please check back shortly for the archive.
All right, we're going to go ahead and get started. I'm Stephen Willey, one of the senior biotech analysts here at Stifel, and glad to have with us for the next discussion the CEO of Pyxis Oncology, Lara Sullivan. Lara, thanks for coming.
Thank you for having me.
Maybe we can start with a quick overview of the company, kind of who you are, what you do, and then we'll jump right into some Q&A.
Pyxis Oncology is an ADC-focused company. We have a lead asset in clinical development, which is a novel target addressing a target called extracellular domain B, or EDB. It's a splice variant of fibronectin, and it's a noncellular target in the tumor microenvironment. Our ADC is an extracellular cleaving ADC. The ADC addresses the target in the tumor microenvironment, cleaves in that extracellular matrix, and then begins to kill tumors through a three-prong mechanism: direct tumor killing, bystander effect, and immunogenic cell death. We brought this asset into the clinic in early 2023 and read out our dose escalation data of a basket trial last fall, last November. We showed tumor regression in six tumor types with a clear, strong signal of a confirmed ORR of 50% in head and neck. We started our expansion trials at the 5.4 mg/kg dose earlier this year in monotherapy.
We also started a combo therapy trial in combo with Pembro.
All right, very good. You mentioned MICVO, though. It's just kind of the centerpiece of the story here. Obviously, the first in concept approach to ADC design and payload delivery. Maybe you can just talk a little bit about it. I think you referenced this a little bit in your intro comment, but what kind of specifically differentiates this type of ADC from a design and mechanistic perspective? What do you have from a translational data generation perspective that supports that?
Yeah, so the clinical ADCs that are in development and have been approved have taken the canonical approach that's been in practice over the last really several decades of being cell surface targeting ADCs, meaning the assumption had been that for an ADC to work, you have to.