Autolus Therapeutics plc (AUTL) Earnings Call Transcript & Summary

Hello, ladies and gentlemen, and welcome to the Autolus Therapeutics AACR II Virtual Meeting Data Update. As a reminder, this conference call is being recorded. I would now like to turn the conference over to your host, Dr. Lucinda Crabtree, Vice President, Head of Investor Relations. Please go ahead.

Thank you, Jill. Turning to Slide 2. Good morning or good afternoon, everyone, and thank you for taking part in today's call on the AACR II Virtual Meeting Data Update. I am Lucinda Crabtree, Vice President and Head of Investor Relations. With me today are Dr. Christian Itin, our Chairman and Chief Executive Officer; Dr. Martin Pulé, our Founder and Chief Scientific Officer; Dr. Vijay Reddy, our Chief Medical Officer; Dr. Muhammad Al-Hajj, our Head of Translational Medicine; Dr. Marco Della Peruta, a senior scientist in the immunobiology department of our R&D team; Dr. Mathieu Ferrari, an Associate Director of Binders Discovery and our R&D team. Before we begin, I would like to remind you that during this call, we will be making forward-looking statements. All statements other than statements of historical facts contained in this presentation are forward-looking statements. Our actual results, performance or achievements may be materially different from those expressed or implied by the forward-looking statements. For a discussion of the risks and uncertainties related to our business and other important factors, any of which could cause our actual results to differ from those contained in the forward-looking statements, please see the section entitled Risk Factors in our annual reports on Form 20-F filed on March 3, 2020, as well as discussions of potential risks, uncertainties and other important factors in our other periodic filings with the SEC. The forward-looking statements contained in this presentation reflect the company's views as of the date of this presentation regarding future events, and the company does not assume any obligation to update any forward-looking statement. You should, therefore, not rely on these forward-looking statements as representing the company's views as of any subsequent date to the present -- or any date subsequent to the date of this presentation. On Slide 3, you will see the agenda for today, and it is as follows: Christian will provide a brief introduction; and that will be followed by a review of the AUTO5 data by Mathieu; after which Mo will provide an overview of the AUTO6NG data; followed by a review of the AUTO7 data by Marco. We will then hand back to Dr. Christian Itin, who will summarize the key messages for today, and provide concluding comments. And of course, we will welcome your questions following our remarks, which will be taken by the panel outlined. So with that, I'd like to turn you over to Christian for the introduction. Thank you. Turning over to Christian.

Thank you very much. Well, welcome, everyone. It's exciting to welcome you to our update on the AACR presentations that were held this week. We're going to be presenting data from 3 of our preclinical programs, and are excited to move forward with the presentation here. So first of all, I'd like to briefly introduce you to the speakers today, in addition to Martin Pulé and Vijay Peddareddigari, who are both well-known to most of you; Martin, a Founder of the company and Chief Scientific Officer; and Vijay, the Chief Medical Officer of the company. We're going to be joined today by Muhammed Al-Hajj, who runs our translational medicine team, who will give you an update on the AUTO6NG program. We'll hear from Marco Della Peruta on the AUTO7 program, first time actually presenting data on our prostate cancer program. And we're going to be joined by Mathieu Ferrari on the AUTO5 program, which is, obviously, targeting T-cell lymphoma, a program we're very excited about. So when we have a quick look on the next slide, #7 (sic) [ #6 ], this is a quick summary of our early stage programs. And you'll see we're going to be presenting on 3 of our 6 programs that we're public about at this point in time. And all of these programs are moving forward towards clinical activity during the course of next year. We'll move with this to Slide #7. And what I would like to do on Slide #7 is just briefly remind you of the basic structure that we're using to program our T cells. What we're doing is we're using the protein modules that change the properties of those cells. And they can be in the category of targeting devices, elements of control, enhancing the activity and allowing the cells to adapt to complex microenvironments, the challenging surroundings that we find around tumor cells, in particular, but also modules that can actually lead us towards an allogeneic phenotype as well. So the way we express these modules is, in fact, typically as a single open reading frame driven off a single promoter, and it gives us a very easy, very straightforward way to actually introduce the information using viral vectors, and actually create complex cell programming with those modules. In fact, you'll hear in the presentations from Muhammad Al-Hajj as well as from Marco Della Peruta, you'll hear about our programs, AUTO6NG and AUTO7, that use several modules to actually address the challenges that we have with regards to the programming and of sales that actually are in a difficult environment, particularly in solid tumor settings. So with that, what I'd like to do is actually hand over on Slide #8, to Mathieu Ferrari, who will introduce us to the AUTO5 program for the treatment of T-cell lymphoma. Mathieu?

Thank you, Christian, and good morning, everyone. I'm very pleased to begin this presentation by taking you through the AUTO5 data update that we have provided at this year's AACR conference on our T-cell lymphoma program. So turning to Slide 9, please. First is some background to the disease or subset of diseases that constitute T-cell lymphoma. They represent approximately 10% to 20% of all non-Hodgkin's lymphoma, and typically manifest aggressively with poor prognosis for patients. Standard of care is variable, and often based on high-dose chemotherapy followed by stem cell transplantation. However, a large portion of T-cell lymphoma patients relapse following treatment with standard therapies. And unlike concomitant B cell malignancies, the T-cell lymphomas haven't benefited from advances in the immunotherapeutic approaches. And this is mainly due to toxicity associated with pan T cell ablation. Our goal and the challenge of this project is to identify a target that will allow us to eliminate T-cell lymphomas without killing all healthy T cells. So turning to Slide 10. Currently, our program against T-cell lymphomas are all based around 3 key elements: the AUTO4, TRBC1 CAR T cell therapy aimed at targeting TRBC1+ patients, around 40% of all T cell lymphoma population; the AUTO5 TRBC2 CAR T cell therapy aimed at targeting the TRBC2 population, which comprises the remaining 60%. These approaches are bridged by our companion diagnostic test for patient stratification. Turning to Slide 11. An often overlooked feature of the T cell receptor beta chain is that an [ ablation ] chained application has led to the presence of 2 distinct TCR beta constant regions, namely TRBC1 and TRBC2, which provide attractive targets for T cell lymphoma. However, a major challenge in developing a biological targeting drug is the high homology between TRBC1 and TRBC2. When looking at the 2 proteins at an amino acid level, they are identical, say for 4 positions from the protein. And the only accessible difference is a single amino acid inversion, a position 4 and 5, where an [ histology ] viral motif in TRBC1 is inverted to liso asparaginase in TRBC2, as shown on the right-hand side panel. Now turning to Slide 12. The AUTO5 program represents the TRBC2 targeting arm of the T cell lymphoma approach developed by Autolus. Turning to Slide 13. We were able to obtain our crystal structure of our humanized anti-TRBC1 antibody complex with TRBC1 to a high-resolution of 2.4-ångströms. On the figure on the left, the construction of the full T cell complex shows binding of the anti-TRBC1 antibody to the TCR beta constant region in close proximity to the CD3 epsilon domain of the TCR complex. As shown in more detail on the right-hand side panel, upon closer inspection of the complex, it becomes evident that the anti-TRBC1 antibody makes relatively small contact with the beta chain. And primarily interacting with residues within the CDR1 and CDR3 of the heavy chain, pinching the 2 residues, asparaginase and lysine, they are important for the TRBC1 and TRBC2 distinction. Turning to Slide 14. Using a structure-guided molecular modeling and in silico mutagenesis, we identified key contact residues within the CDR1 and CDR3 of the antibodies, that when mutated, allows for a change of specificity from TRBC1 to TRBC2. And here, we show a progressive mutational panel, including 1, 2 or 3 simultaneous mutations on panels D, E and F, particularly increased binding for TRBC2 and combinating with mutant 3 in green on the last panel, number F, showing an inverted specificity compared to the parental anti-TRBC1 clone, which is marked in red. So turning to Slide 15. The newly developed anti-TRBC2 binder was then incorporated in our second-generation CAR with different spacer and endodomain confirmation to improve killing response. And the optimized architecture is showing here an efficient and specific killing of TRBC2 positive PBMCs from healthy donors. Turning to Slide 16. Alongside the killing efficiency, the optimized anti-TRBC2 CAR T cells are also able to specifically proliferate only in response to interaction with TRBC2 target cells, as shown in the last panel. Controlled anti-CD19 in CAR T instead show no response to any of the cell target [ IFN ]. And similarly, the anti-TRBC2 CAR T shows strong and specific cytokine release response for an interferon gamma and IL-2, one cocultured with TRBC2+ target cells, while no significant cytokine release was associated to culturing with TRBC1 or TCR local target cells. So turning to Slide 17. The same specific effect, and was also served in vivo, when the mice were challenged in a co-infused mixed TRBC1, TRBC2 tumor model. The anti-TRBC2 CAR was able to clear the TRBC2 cells, while sparing the TRBC1 population, while our anti-TRBC1, the AUTO4 construct, showed the opposite effect. No effect was observed with non-targeting control, anti-CD19 CAR T. This highlights once again the specificity and selectivity of both arms of our T cell lymphoma product. Turning to Slide 18. In summary, we have demonstrated that targeting the TCR beta constant chain isoform can provide a generic approach to [ tackle the multiple subtypes ] of T cell lymphoma. The AUTO4 program, targeting the TRBC1-expressing lymphoma, is currently in clinical evaluation. Here, we have described the development of the AUTO5 component of our T cell program, targeting TRBC2 expressing lymphomas. We have shown the discovery of our selective anti-TRBC2 binders for CAR T cell development. Further, our preclinical data package has demonstrated the utility of the anti-TRBC2 CAR both in vitro and in vivo. And we have shown efficient and specific CAR T proliferation upon targets and interaction in vitro, and demonstrate a significant cytokine release in response to target cell counter. The in vivo clearance of TRBC2 tumor in a mixed tumor model using NSG mice further highlights the specificity and selectivity of our approach. So turning to Slide 19. I will leave then now the stage to the next presenter.

Thank you, Matthew, and good morning, everyone. Pleased to have the opportunity to walk you through the AUTO6NG data update that we have provided this year at the AACR conference, where we focus on the -- on our ability to overcome the immune suppressive mechanisms in the tumor micro environment and demonstrated preclinical antitumor activity in GD2-expressing solid tumors. Turning to Slide 20. AUTO6 is a GD2-targeting CAR that was designed with a binder that minimizes on-target off-tumor toxicity, and the binder was also humanized to reduce immunogenicity and our RQR8 co-switch was incorporated into the design to ensure safety. A Phase I clinical trial in relapsed/refractory neuroblastoma was conducted by Cancer Research UK and the results which were presented at the 2018 AACR demonstrated clinical activity as evident in the decrease of disease hot spots by MIBG scans following AUTO6 treatment. In that trial, no dose-limiting tox and no neuro tox were observed. Now our strategy is to derisk the target and the CAR in the neuroblastoma setting, where GD2 expression is high and homogeneous. And now with this initial clinical proof of concept, we proceeded to explore the applicability of this GD2-targeting CAR in large indications where the target is the reported to be [ expressed ]. Turning to Slide 21. We started by analyzing its microarray set of multiple solid tumors that identified small cell lung cancer as one where GD2 appears to be expressed at high levels in the subset of patients. We then look closely at the GD2 expression in the panel of 25 primary human small cell lung cancer section. And here, we found about 68% of these patients' biopsies expressed GD2, while about 32% did not express it. Following this confirmation of expression in a significant percentage of SCLC biopsies, we decided to functionally evaluate targeting GD2-expressing small cell lung cancer cells using the H446 cell line model with AUTO6, both in vitro and in vivo. Turning to Slide 22. Now outside of a relevant target in the tumor type there are other key elements at play that could impact the CAR T therapeutic approach. It's established that multiple pathways are deployed by the tumor to evade the host immune system, in particular, TGFBeta signaling, which is an essential pathway that plays a multiplicity of roles in shaping the tumor invasion mechanisms, in addition to checkpoints, which consist of inhibitory and stimulatory elements that also impact the antitumor immune response. So we set out to assess the status of this element in a small cell lung cancer setting to inform on which modules may or may not be required to elicit efficacy in that indication. Next slide, please. So to do that, we designed a comprehensive panel of biological markets to characterize cancer cells and the cells in their [ rigidity ] and their phenotypic state. And we deployed tissue cytometry to conduct a multiplex analysis of small cell lung cancer patient biopsies for an in-depth evaluation. So now one of the findings was the high level of phospho-Smad3 expression, which is a downstream TGF-beta signal on the left panel. Mostly in the stroma, and to a lesser extent in the tumor cells themselves, suggesting a need for the T cells to overcome such an immune suppressing mechanism in this indication, to bypass the signal and reach the tumor cells. The middle panel reflects the diversity of the cell types present in the microenvironment, and the right panel shows, among other things, a negative or weak signal of PD-L1 in this biopsy, not surprising given the therapy-naive nature of these biopsies. But we expect that PD-L1 will likely be up-regulated once T cells come into play and engage with the tumor. Turning to Slide 24. So we started with testing the impact of the CAR T cells containing a GD2 CAR, 41BB-Zeta signaling domain and the RQR8 safety switch in the H446 cell line expressing GD2. The left panel shows that the AUTO6 CAR Ts are readily able to eliminate the H446 cells in vitro after 48-hour coculture. However, when we tested the same tumor model established in vivo this time, we observed no difference in tumor growth between the mice treated with AUTO6 or the one, the nontransduced control T cells, indicating no efficacy in that setting. Turning to Slide 25. So to investigate whether the lack of in vivo efficacy was due to insufficient homing of AUTO6 to the tumor site, we generated AUTO6 cells that coexpress luciferase to track the bioluminescence in mice with established H446 tumors. On the top row, you see the luminescent signal disappearing in the mice that received control T cells expressing luciferase only at the tumor site. AUTO6 CAR T cells, on the other hand, co-localized and persisted at the tumor site, the bottom row. So these observations suggested that the inability of AUTO6 cells to control tumor growth in vivo was not due to inability to hone to the tumor site, but rather they were secondary to other inhibitory mechanisms. Turning to Slide 26. So to address this initial lack of efficacy in the in vivo setting, we move to test whether enhancing modules designed to overcome 2 of the key immuno-suppressing mechanisms in vivo that I described earlier, TGFBeta end checkpoint, along with the cytokine signal module to enhance CAR T persistence will lead to an efficacious outcome. Autolus has developed a series of core enhancing modules, 3 of which are capable of tackling the 3 elements that Christian alluded to earlier. In addition to the GD2 CAR and the RQR8 safety switch, we [indiscernible] the impact of adding a dominant negative SHP2 module, which endows checkpoints pathway resistance, any down negative TGF-beta module, which shields CAR Ts [ from immuno- ] suppressive signals and the cytokine module, which is an IL-7 separate chimeric protein that improves CAR T persistence. Turning to Slide 27. So we generated 2 next-generation AUTO6 version, one version co-expressing the anti-GD2 CAR, along with the dominant negative SHP2 and the down negative TGFBeta, and the other version co-expressing all the elements on deferred surge in addition to a CCR7 module incorporation. Then we tested 2 versions functionally in vitro. As you can see from the panel to the right, both version are equally effective at eliminating the H446 target cells in vitro at the various efficacy [ tolerance ] ratios. Slide 28, please. And then we tested the functionality of these 2 versions in vivo with established H446 tumors. Here, in contrast with the earlier observation with AUTO6 CAR alone, which is at Slide 23, we observed that both AUTO6NG versions were highly efficacious and effective in controlling tumor growth in vivo. This suggested that overcoming a TGFBeta end checkpoint's immune-suppressive signals are required to induce efficacy in this small cell lung cancer model in vivo, and a simple CAR targeting GD2 would not control tumor growth. Turning to Slide 29. To evaluate whether the CCR module -- CCR7 module impacted expansion and survival of the CAR T cells, we quantified the number of tumor-infiltrating CAR Ts at the tumor implantation sites and at the spleen at the time of euthanasia. This was done by flow cytometry, and both the AUTO6NG versions were detected at this late time point. However, there were more CAR T cells from the AUTO6NG expressing the CCR7 modules at the tumor infiltration site and at the spleen versus the one without it, suggesting that the module had an enhancing impact on the CAR T expansion [ emphasis ]. Slide 30, please. So potential risk from incorporating a CCR7 module is increased level of cytokine as a result. To assess the impact of the incorporated CCR module, we quantified for inflammatory cytokines in mouse sera collected at day 34 after CAR T cell infusion. Higher level of interferon gamma, as you can see in TNF alpha and GM-CSF, were detected in mice treated with either version of AUTO6NG relative to the non-transduced controls, which was expected given the efficacy observed in these mice. However, there were no significant difference between the 2 or 3 AUTO6NG versions groups, suggesting that although CCR7 increased CAR T expansion persistence, it does not appear to drive higher level of cytokines. Slide 31, please. So in conclusion, AUTO6 is a GD2-targeting CAR that is clinically active in neuroblastoma. Our experimental data indicates that GD2 is also an attractive small cell lung cancer CAR T target. Here we have evidence -- we showed evidence that AUTO6 alone is not sufficient to drive in vivo efficacy in its small cell lung cancer mouse model and additional cell programming modules rendering the CAR T insensitive to TGFBeta signaling and checkpoint inhibition are required to drive efficacy. Also showed that the modules enabled the AUTO6NG CARs to persist and be detected by flow on day 34 post infusion and the expansion enhancement was without increased cytokine levels in vivo. AUTO6 will be clinically explored NG -- the AUTO6NG will be clinically explored in GD2-expressing tumors. Turning to Slide 32.

Thank you, Muhammad, and good morning, everyone. I'm very pleased to have the opportunity to walk you through the AUTO7 data update. We have provided a slide to your conference to provide the update on our Anti-PSMA humanized CAR T cells with improved persistence and resistence to the tumor macro environment for metastatic castration-resistant prostate cancer. Turning to Slide 33. Prostate cancer is the second most common male cancer worldwide with highly unmet medical need, patients diagnosed with localized disease are usually treated with surgery or radiotherapy. Yes, many develop more advanced recurrent disease and patients with advanced prostate cancer ultimately progress to become especially resistant followed by metastatic disease. There is no existing curative treatment. However, site-based therapies have the potential to be curative. Turning to Slide 34. To develop a successful treatment for advanced prostate cancer, there are various biological hurdles to address the complex tumor macroenvironment. This is characterized by low number of T cell infiltrates, high presence of TGFBeta and PD-1, PD-L1 immunotherapy pathways and low PH values. In prostate cancer, there is not one single challenge to overcome, but other several challenges that need to be addressed. Turning to Slide 35. AUTO7 is a multi modular CAR T cell program to generate resilient CAR T cells withstanding hostile tumor environment by introducing model to affect the feature tumor biology. We choose PSMA2 tumor CCRT antigen, a specific target for our products, and developed a novel anti-PSMA binder, engineer our PSMA CAR module. We introduced a dSHP2 module able to shield AUTO7 from checkpoint in vitro. A dominant negative TGFBeta receptor 2, able to work as decoy for the TGFBeta sequence, a chimeric cytokine receptor to support CAR T cell survival; and finally, a model that activates immune response at the tumor site through limited secretion of [ cytokines ]. All models effects are limited to the CAR T cells and limited surrounding environment rather than having a systemic impact. Turning to Slide 36. AUTO7 is a genetically engineered product designed to tackle a complex tumor environment. AUTO7 is characterized by CAR T cells expressed in all these modules to create a resilient product. All AUTO7 models are delivered by use of gamma-retroviral vector system. Turning to Slide 37. We developed AUTO7 CAR program to target the PSMA tumor-assisted antigen. PSMA is one of the most commonly expressed genes found in prostate cancer, and each expression can vary with the degree of differentiation. Our first aim was to design a sensitive CAR. Turning to Slide 38. We selected and humanized our own PSMA binder in [ indiscernible ] 7A12. This displayed identical staining profile to the well clinically characterized anti-PSMA binder J591. By a physical characterization of the 2 binders revealed similar affinities with overlapping [ epitopes ]. However, our band was found to be more stable than J591 across a range of PH values expected to be in the tumor macroenvironment. Turning to Slide 39. In the CAR format, 7A12 is characterized by [ P shift ] specific cytotoxicity against cell lines that have both low and high level of PSMA antigen. AUTO7 CAR was able to lie in cell lines expressing low levels of PSMA, less than 600 copies, as effectively as target cells expressing more than 16,000 copies of PSMA. Turning to Slide 40. TGFBeta is an anti-inflammatory cytokine and is expressed at high levels in prostate cancer within the tumor microenvironment. Now binding to these receptors 1 and 2 on T cells, to just be [ compromised ] by immunosuppressive effects. Turning to Slide 41. The dominant negative TGFBeta receptor 2 blocks the intracellar domain and works as decoy by preventing the activation of immunosuppressive [ response ], similarly, the usual course of the TGFBeta binding. We coculture PSMA-expressing target cells with T cells expressing AUTO7, or anti-PSMA CAR only, in presence or absence TGFBeta and normalized to introduce human [indiscernible] in the same conditions. In CAR-only tumors, the specific cytotoxicity against PSMA-positive target site, green on the left side of the graph, was reduced by presence of TGFBeta, in purple. In AUTO7 CAR T cells, the dominant negative TGFBeta receptor 2 model makes CAR T cells resistant to inhibition and [indiscernible] of data. As you can see in the [ triangles ] on the right side. Turning to Slide 42. Inhibitory proteins expressed by the tumor cell, in particular, PD-L1, in prior CAR T function, the dSHP2 model takes advantage of the fact that many of these inhibitory proteins which allows the downstream phospho-dSHP2. Turning to Slide 43. dSHP2 module expresses a truncated inactive form of SHP2 that competes with endogenous SHP2 for value to mutate the effectors like PD-1. This blocks the signal from these inhibitory checkpoint proteins shielding CAR T cells from inhibition. Unlike monoclonal checkpoint blockade, dSHP2 is restricted to the CAR T cells and can block multiple inhibitory receptors. We over-expressed PD-L1 in CAR only and AUTO7 CAR T cells, a coculture with PSMA-positive target sites, and we target 5 co-expressed PSMA and PD-L1. CAR-only samples showed reduced killing of PD-L1 expressing target cells when compared to target cells with our PD-L1 overexpression. PD-L1 over-expression in target therapy do not interfere with the cytolytic activity of AUTO7 ensuring that dSHP2 model maintains AUTO7 cytolytic function in the presence of PD-1/PD-L1 signal. Turning to Slide 44. It is well-known that CAR T cell persistence correlates with clinical efficiency, and lack of persistence result in poorer performance of CAR T cell program in solid tumors. We have developed the IL-7 CCR module that delivers a consistently active cytokine receptor. Turning to Slide 45. IL-7 CCR module is based on a substructure fused to the I7 receptor alpha and chromo-gamma [ chain ] receptor. Its functionality in all the cells in AUTO7 by separating [indiscernible] of exogenous cytokines at 7 days, and compared to CAR only and AUTO7, now we know the IL-7 CCR module [ signal for control ]. Both of which should pull proliferation, while AUTO7 showed increased proliferation in use by the IL-7 CCR model. We then challenge CAR only and AUTO7 to free route of CDF killing by weekly addition of target cells and measure tumor cell survival. With the CAR only, we reported a drop in activity within the second week of the simulation with consequent reduction in [indiscernible] orally. In comparison, AUTO7 maintained prolonged cytolytic activity and cytokine suppression, confirming that I7 CCR module, not only extend AUTO7 T cells persistence, but extends AUTO7 activity through multiple rounds of restimulation.. Turning to Slide 46. IL-12 is a pleiotropic cytokine that which creates connection between their innate and adaptive immunity. IL-12 has a potent antitumor activity, but I levels that has been associated with systemic toxicity. Turning to Slide 45. Our [ subscale ] IL-12 module has been developed to control the expression of IL-12 to prevent toxicity by drastically reducing IL-12 secretion while still maintain potent adjuvant activity. This module has been developed to turn an immunologically cold tumor into alpha. The graph [ at the right ] display IL-12 secretion by T cells that use recurring method, PD-2 and peptide, during the lifetime [ penetrate ] scheme or other sub scale IL-12 module. This top scape IL-12 module is detectable but drastically reduced IL-12 expression when compared to the [ y-type IL7 ] sequence. Turning to Slide 48. Stop skip module functionality was tested in vivo by comparing CAR stop skip IL-12 to CAR-2A-IL12 in a model of B16 F10 target cells in immunocompenent mice. Mice had received target cells on day 0 and T cells on day 7. On the right side of the slide, we report the drastic weight loss measure in the CAR-2A-IL12 group to IL-12 systemic toxicity. In comparison, CAR stop skip IL-12 group did not show weight loss. On the right side of the slide, we can see that CAR only has shown failure to control tumor growth. While mice infused with CAR stop skip IL-12 showed reduction and control of tumor growth, while preventing toxicity. Turning to Slide 49. Lastly, we tested AUTO7 in a xenograft model of human prostate cancer. Turning to Slide 50. NSG mice were injected in the flank with 5 million PC3 prostate cancer [ human ] cells translates to expressed PSMA. Followed 3 weeks after when the tumors were palpable, by injection of 1 million AUTO7 cells. We compare our AUTO7 cohort to CAR only and tranduces PC3-PSMA as control. While CAR only was not able to control the tumor growth, AUTO7 cells completely eradicated the PSMA-positive site without toxicity. Turning to Slide 51. In conclusion, our results demonstrate the feasibility and efficacy of the multimodular AUTO7 product designed to overcome immunotherapeutic challenges presented by the advanced prostate cancer. In particular, AUTO7 is valued importantly in cytotoxicity assays against cells expressing PSMA even now to lower levels. dSHP2 models desensitizes AUTO7 to PD-1/PDL-1 mediated immunosuppression. The dominant negativity TGFBeta receptor 2 shields AUTO7 from TGFBeta inhibitory effect. The IL-7 CCR module confirms AUTO7 cytokine independent proliferation and improves the ability to serially kill targets. And last, the stop skip 12 module cells within AUTO7 contains potent adjuvant activity with reduced risk of toxicity. AUTO7 models altogether announce T cell function by producing releases to TGFBeta and PD-1/PDL-1-driven inhibition by extending persistence, proliferation and activation. I would like to thank the team involving AUTO7 development, and thank you for your attention. I will now pass the call back to Dr. Christian Itin.

Thank you very much, Marco. Much appreciated. And moving on to the next slide, please. What we're looking to do with the programs to -- developing at Autolus is that we're looking to tailor each one of the programs to the particular disease entity that we have in front of us, and we're looking to tackle. If we go through our current pipeline, and starting out with our lead program, most advanced program, AUTO1, which we're currently moving in a pivotal study for the treatment of patients with an acute lymphoblastic leukemia, we designed a product there that is particularly well suited for that disease. What we needed to have was very high level of activity, an ability to put pressure on the tumor for an extremely long period of time, so excellent persistence, and we also needed a good safety profile because we're dealing with patients that are significantly beaten up and tend to be very frail. So the design there was to actually have these 3 key properties as the primary objective for the product, and that resulted in AUTO1. The second program we're running is for the patients with diffuse large B-cell lymphoma. This is a patient group where, again, you need a high level of clinical activity. And you need to particularly make sure that the complete remissions you induce, you can maintain and what we basically focused on with the program is to shut down 2 routes of escape for the lymphoma. One is related to antigen loss, the other one related to checkpoint upregulation. The design we've chosen with this product is a dual targeting approach that give us an ability to really tackle the particular challenges in this disease setting. And as we're going from there to our AUTO4 program, the sister program to AUTO5, with these 2 programs, it's all about having a very sophisticated way of targeting that gives you an ability to target the actual malignant cells without targeting at the same time all T-cells in the patient's body so that you can actually preserve T-cell immunity, while at the same time, you can tackle the particular disease. And as we move from there, you have then the examples with AUTO6NG and the example on AUTO7 where we're starting to deal with solid tumor settings that have a quite challenging environment that they create around themselves and that need to be tackled in a way that allows us to, on the one hand, enhance the resilience of the CAR T cells against these defense mechanisms that the tumor is using. We're looking at the family of checkpoints, TGFBeta, but also the fact that we need to make sure that the CAR T-cells can actually survive and persist in a very difficult environment as well, and that is where we're adding the additional module of the cytokine module. Now we extended that then one step further, in what's typically referred to as an immunologically cold tumor in prostate cancer and by actually secreting a cytokine, allows us actually to shift that in vivo into one that is actually supportive for the immune system. And with that, actually, create a tumor environment where the immune system per se can be highly active, and that is a critical piece. The technical challenges, Marco walked you through there is that if you actually secrete a large amount of cytokine, it becomes highly toxic, and it is actually not easily tolerable. So what you have to find is a way to actually have a high local concentration in the tumor without inducing systemic toxicity, and that is what we accomplished with a particular technology developed there. So with that, what I'd like to do is actually open up for questions and I'm looking forward to the interaction.

[Operator Instructions] Our first question comes from Mara Goldstein with Mizuho.

Just a question on the TRBC targeting program. And within the data that you showed around cytokine targeting, is there a minimum threshold that we should be thinking about that translates from the activation you saw in preclinical models into human models? And if you could just provide an update for companion diagnostics for AUTO4 and AUTO5, that would be great.

Right. So I'll start out with the targeting component. Obviously, the targeting has to be very exquisite and highly specific. And this is what Mathieu showed you, in particular, with the panels where both cell types were infused into a mouse model treated, and you wanted to see that one of the subsets survive versus the other one actually is taken out. And I think what we could see in those experiments is an incredible level of specificity. With regards to the companion diagnostic, which also relates to the T-cell lymphoma program, we've actually -- at a point that we actually do have a companion diagnostic that allows us to identify the patient tissue and determine whether or not that tumor is TRBC1 or TRBC2 positive, and we're using that in our clinical trial to identify the patients. And that companion diagnostic obviously gets further validated as we're running through the clinical study. The third aspect of the question that you asked was actually related to the amount of cytokine released that we need to have locally. And one of the remarkable things is obviously that you can have a very low level of secretion from the T-cell because the extra cellular space that is actually present in tumors are a very finite, very small space. So that still, you'd be able to actually get a high level, a high local level of concentration of the cytokine without inducing toxicity. And I think the experiment that had demonstrated it very nicely was the in vivo experiment where we could actually show that. Normal IL-12 secretion, which you just normally produce it at the regular levels that you would expect in a -- just from a normal promoter. You did see that, in fact, the animals had to be terminated within a few days because, in fact, they were actually having massive cytokine release and corresponding adverse events so the animals had to be terminated. Whereas with the very low level of expression that we're providing in the actual drug product, we can avoid that level of toxicity. And at the same time, obviously, get full activity in the tumor as is demonstrated in the model.

Our next question comes from Jim Birchenough with Wells Fargo.

I guess for AUTO7, maybe just thoughts on what the therapeutic hurdle would be? Do you view the PSMA directed radionuclide therapeutics is setting the bar? Or how do you think about the level of efficacy you're looking to target there? And then I have a follow-up.

Yes. So obviously, it's a field where we have seen a lot of activity in the back end of the disease. And there are a relatively small number of agents in terms of single-agent activity that show a certain level, although relatively small activity. I think there's a number of combinations that we see active there. And what I would like to do is actually hand over to Vijay to give a brief view on the landscape in prostate cancer.

Thanks, Christian. As you know, prostate cancer patients, once they exhaust their option for androgen deprivation therapies, go into chemotherapy. And after chemotherapy, basically, there is no standard of care. And that's a space where the most new drugs are being developed. As recent immunotherapy that was evaluated there with [ NEMO ] lipid combination had a response rate of 10%. So almost nothing works there when it's quite cold. And with regards to radio -- our new clients, most of them induce a PSA response. They don't actually induce a resist or actually too much in case so all this is a biomarker response. So the hurdle to show an efficacious therapy, such as with the CAR T, is quite low. That said, if you have a -- actual tumors [ engaged ] and not just a biomarker response, which is in the range of 30% to 50% with complete responses as well, you would have a very active treatment for a solid tumor.

And then maybe just a broader question, Christian. You have the ability to generate multiple or to deliver multiple modules with your vectors. What is the ultimate capacity, number one, to deliver different payloads? And then just as you think about next-generation therapeutics here, how does your translational work guide what you've designed into these modules? Because you're doing a number of different things, dual antigen targeting, varying the binder, trying to address the tumor microenvironment. So how does the translational work inform what goes into those modules and what's the ultimate capacity?

Right. I'll start with the second question in terms of how we actually inform kind of our decision, what to program for. I think a lot comes down to looking at the actual tumor tissue and develop an understanding ourselves as well as what others do in the field, to develop an understanding of what are the mechanisms that work in a particular tumor type. And I think if you just think back on kind of the images that we've all remember on PD-1 impact as an example, and we all recognize that even if you have a response with the PD-1 monoclonal antibody, it is rarely a complete response. What you do see is you do have shrinkage rates of various degrees in various stations, at various stations throughout the patient body. And what that indicates to you is that there is an element or there's a group of cells within that patient that are particularly well protected by checkpoint. Once you lift that protection, you can actually get to the target cell and can remove it. The fact that there actually a still tumor remains untouched after the therapy tells you that while checkpoint was relevant, for a subset of those cells, there are other mechanisms that actually held the activity back. And that is why we believe it is important to actually provide the CAR T cell with an ability to have intrinsic resilience against a range of defense mechanisms that the tumor can deploy. And we do believe, as you can see, with the 2 programs is that a combination of addressing the inhibition through the family of checkpoints, not just PD-1, PD-L1, but the family of checkpoints. And on the other hand, addressing TGFBeta, in combination with the survival signal, should get us a long way to where we need to be. And then in indications like prostate cancer, as Vijay pointed out, which tend to be immunologically cold, to then actually turn that into an immunologically hot tumor by actually inducing -- and inducing that through a locally secreted cytokine. We believe we can actually change that microenvironment in a very significant way. Obviously, we're looking much more deeply into these tumors. There are obviously many markers that Hamad and his team actually have been analyzing and that we're looking at. And we're looking at that part of the additional biology and are currently determining whether indeed there is a need for additional ability to sort of cope with some of these additional mechanisms that we start to pick up as well. But that's an ongoing process, and we're moving through that. In terms of the amount of programming that we can actually include and deliver in a vector, I'd like actually -- to hand over to Martin to maybe briefly give us a sense for kind of the payload we can deliver with the vector systems.

It's Martin here. So a gamma retroviral vector can usually take around 8,000 base first wave kilobases. Our components -- most of our components are quite small. They range in coding sequence lengths from around 500 to 1.5 -- 500 basis, 1.5 kb. So we can comfortably get in quite a lot. But then I think you'll have seen in earlier presentations, what we're starting to do is to actually have double transductions. So we have 2 vectors transducing the T cells at the same time. And if you do that, you really have quite a lot of space to play with. So at the moment, vector capacity is really not a limitation for us.

Our next question comes from Aaron Welch with H.C. Wainwright.

Thanks for the data update. So it sounds like you're -- if you're looking for resist responses in prostate cancer, you might be trying to avoid patients with only bony disease or primarily bony disease? So is this a strategy that you're going for? And do you expect bony metastases to present any exceptional difficulties in treating the prostate cancer patients? And I have some follow-ups.

It's a very good question, Aaron. And I think at this point, we're probably all short of data with regards to the ability of T cell media therapies to home in and take on bone mets. And I think that is going to be one of the questions that we'll have to actually investigate in the clinical development of the program. I think it's one of the places where we're sort of operating in the absence of data. We do know, obviously, that bone impact is certainly not only -- or an issue with prostate cancer as an example, but it can even be an issue or a challenge even in hematological cancers, in particular, multiple myeloma, we have remodel -- modeling of bone and changes in bone structure and compartments that you sort of create quite unusual microenvironments. So it is an area that we've, obviously, have encounter with other disease settings. It does look like in the multiple myeloma field that we seem to be active pretty much across, but to what extent and whether that extends all the way through to the type of bone mets that we're seeing with prostate cancer remains to be seen. I don't think we have the answer at this point.

Okay. So I assume that you guys don't have any mouse models of bony disease. And then I also wanted to ask about what's the timing of getting AUTO7 into the clinic?

Yes. So what we're planning to do with AUTO7 is, obviously, that we would like to get this program into the clinic during the course of next year. In terms of good models, I'm happy to sort of hand over to [ Mo ] on that, but to my knowledge, and not sure that there are really predictive models for bone mets in mice that would be suitable for the type of approach that we're doing. But maybe Muhammad can expand on this a bit.

No. Actually, I don't have a lot to add. It's true. The models, they're not -- they're available, really not that predictive. And the best case is basically to design it the best way you think it's going to work and to test it out in patients.

I'm sorry, this is Vijay. PSMA-PET is fast becoming standard of care. And that will simplify assessment of bony disease.

Our next question comes from Chad Messer with Needham.

Great. And thank you all for a very exciting presentation this morning on your data coming out of AACR. A few, if I may, back on the IL-12 module, sort of the newest little bell and whistle you're putting on here. First, and I apologize, maybe I just missed this, but it wasn't clear. In the experiment where you showed that IL-12 was required to get efficacy, was one of the controls, specifically the same AUTO7 minus IL-12? And then another question on it, and we didn't really get into the mechanics of how you created [ an aside ] a modified model, that lower release. Is that something you were able to dial in? And were you able to kind of optimize the amount or it's just sort of high versus low? And then finally, I'm kind of interested in any work you're doing or you would be interested in doing with this module in other programs?

So I'll start at the back, and then I would like to hand over to Martin. In terms of additional indications, clearly, there are additional indications that I think are kind of suitable and particularly interesting for an addition of an IL-12 module, considering this is quite a range of solid tumor settings that are considered immunologically cold. And I think in all those, this type of a model, I think, would be well suited. But I would like to hand over to Martin to maybe explain how we actually got to a low level expressers and the kind of technical approach that we're taking to do that.

Thank you, Christian. We've borrowed something from sort of immuno-bacterial vectors here. So there's a code on that usually makes the ribosome terminate. So you don't get any further translation but if you flank the stop code unappropriately, the ribosome occasionally keeps on reading through, so then you get translation in the second protein. And then they give the [ delay in ] peptide, the 2 proteins then break in 2. And it's actually very versatile. You can dial down the ratio of expression of the second protein compared to the first protein for a very wide range. We've got quite a lot of versatility, and that depends on the stop code you use and also on the basis that you flank it with. So you can dive it up and you can dial it down. I mean, I think we've picked -- I think the challenge with ISR, which is very toxic systemically and also very toxic when is expressed at high levels from engineered T cells. So we know that's in clinical studies. But the challenge there is to find exactly the right point at which you have the minimum amount of IL-12 being secreted in the microenvironment and that will give you the standard antitherapeutic effect. And I think we've found that. So it is a very, very low level of IL-12, which is really nice. And I think it would be very difficult to achieve this in any other way. So I guess that's the answer to your second question. The answer to your first question. I mean, I think when you do it, whenever -- and this is very standard in the CAR T world, whenever you have a new module, what you do is you take an existing model and usually, you reduce the amount of T cells that you put into the model, so that just sort of your baseline construct doesn't work very well anymore and you see what happens with the -- with addition of the second or additional module.

Okay. Great. Very interesting science. Maybe just 1 follow-up on PSMA. Great target, well-validated in prostate cancer by a number of folks over the years. But you're probably aware, since you're going to be going after late-stage patients, and it can be shed. There's a transformation late in prostate cancer where some patients stop expressing that antigen. Just wondering your plans to mitigate that. I mean, you can maybe scan for it or something like that?

I mean, typically, 1 of the important things with those types of mechanisms is that they're not all are non-mechanisms. And so what you tend to have is quite a range of expression. And Marco, in fact, alluded that -- to that at the beginning of his presentation, and indicated that, that actually was the reason why we wanted to have a chimeric antigen receptor that would have an ability to pick up a wide range, particularly also very low range of expression of PSMA. And that is what we basically designed with this product. In addition, also making sure, which is another mechanism of defense, is that a lot of the solid tumors, actually, when you look at the microenvironment, it's acidified. And the problem with that is that a lot of antibodies actually do lose their binding ability, and you did see that with the control antibody that Marco did actually present. And so the goal here was to actually have an antibody that has a very high level of sensitivity and with that, an ability to actually pick up low quantities of PSMA on the cell's surface. And at the same time, continue to maintain that activity even if you get into an environment that becomes acidified.

Our next question comes from Biren Amin with Jefferies.

I have some questions on the AUTO6NG program. Clearly, you've added the IL-7 module to the NG component. So maybe, Christian, I wanted to ask you, I think on Slide 28, you showed tumor volume reductions with AUTO6NG versus NG, and you're seeing a slight improvement with AUTO6NG. And so, I guess I want to ask you, why wouldn't we see a more significant reduction? And do you think it's -- AUTO6NG is more designed for relapse, so more durable responses? And how do we, I guess, assess that from an animal model? And if you've done those studies, I guess?

The first thing, right, so thanks a lot for joining, Biren. So the first thing, I think, to be said here is remember when Mo started the presentation, he first showed the example of what's happening in this tumor model, if you choose AUTO6 in its current form just for the chimeric antigen receptor. And in fact, what the data showed is that, in fact, the CAR T, although was active in children was not active in this particular tumor model. It could kill individual tumor cells. That was shown in the first experiment, but it could not actually do that once that particular tumor started to form in the animals. So that's the first thing. The second part of the answer is that adding the additional modules clearly translated or transferred this product now in 1 that, starting with the CAR alone that couldn't actually impact the tumor, now the situation where we can impact the tumor. What you can see in the model is, when you look at it is that there is obviously, the range of differentiation that we have to look at with regards to the impact of the CCR7 module on its own. In this particular model, and this particular view, you'd obviously be hard-pressed to show because both are too active. And because of that, there is not enough delta to tease out the difference. But what is important is what you did see actually in the further data that [ Mo ] showed that indeed, there is a significant increase in CAR T cells that we can pick up. And that is one of the important things and one of the real challenges in solid tumors is that you infuse the CAR T cells into the bloodstream, and then they have to actually find the tumor. Normal distribution is, they go part in the lung, they go part of the bone marrow, there's no tumor there. They have to redistribute, and they actually have to find the tumor in tissue. So the number of T-cells that we actually have that are getting there are relatively few and it takes a bit of time for those cells to migrate through all the tissue and actually find the tumor. So the ability of the cells to proliferate and to survive are absolutely crucial in a tumor setting, particularly if the animal gets to speak as a human being. The mouse is still relatively forgiving because it's a relatively small volume, but, obviously, a much bigger challenge when we go into a large being like a human being.

Our next question comes from Matt Phipps with William Blair.

A couple of various questions here. First, Christian, what's the bar with the AUTO4 program that you need to see in order to move AUTO5 into clinics? Is it just response? Do you also want to see durable responses? I know Martin previously presented on a single case patient, I think, responded, but it wasn't too durable?

Matt, first of all, thanks for joining. I think what we want to see is we want to see that we actually have proper biological activity. We want to see that we can actually get a response induced in these patients. And I think that is absolutely critical. Obviously, the initial information that you sort of related to was, obviously, at an extremely low dose level, and we'll have to see, obviously, where we get to and by getting to what we believe to be actual active, fully active dose levels, which we assume are quite a bit above what we have shown initially, which was just a 25 milligram dose. So we expect to actually see proper responses, that's what we want to see. And we actually really want to have both of those programs in, I think, with any sense of -- and any view of an activity, clearly on the size needs to move in the clinic and we need to take it forward.

Great. And then to the IL-12 module, like on Slide 48, you showed the advantage as far as -- mainly, I guess, on the toxicity side, but still maintaining antitumor activity. But have you been able to look in these tumors and see if you're actually getting some signal of endogenous recruitment, maybe if it's just even to bus those stats for, just show signaling of IL-12 or any other marker?

It's an excellent question, and we're absolutely are looking into these questions. One of the challenges with CAR Ts is when they work, they work quickly. And so actually hitting the right time to actually see the various components of the immune system at work is actually not an easy endeavor. And so we're currently investigating into the kind of to what extent we can tractorize this model further to look at the additional impact of immune cells that we can have within those tumors. But one of the important things is there, obviously, also that the models that we have to work with, to be able to demonstrate that have to be wild type mice and not ones that actually have an immune defect, which are typically quite often used in these models. So you have to have a very sophisticated model in the -- to have an ability to see in the mouse itself the ability to recruit to the tumor other immune cells outside of the CAR T cells, but these are investigations that are ongoing.

Okay. And then last question, adding all these modules, particularly things like the IL-7 receptor and IL-12, does that change the phenotype of the cells during manufacturing? Does it lead to any more of a differentiated phenotype just because you're pushing the cells more?

It's a very good question. We don't seem to be picking that up. The way that we're actually doing the programming here, obviously, we're providing an intracellular signal. It's not a classical, full-blown IL-7 signal that we're providing, which is somewhat modified approach that we're using here. And at this point, we haven't seen any significant drift in differentiation state.

Our next question comes from Graig Suvannavejh with Goldman Sachs.

I've got 2 questions. First is on AUTO6NG, there's obviously a range of indications where you can go with your GD2 targeting CAR T candidate. Which indications do you think you'll prioritize, which is whether the smaller ones that maybe you can handle on our own or the larger ones where you might seek to partner? And then my second question just has to do with the AUTO7 program. Very exciting data. It is a crowded space in terms of CAR T efforts in prostate cancer; MustangBio, Poseida, Bellicum, UPenn has got a program. So could you perhaps put in perspective, kind of compare and contrast in terms of where the different programs are and kind of how you think your candidate would be positioned relative to the others?

Well, thanks for joining, Graig. So to your first question, in terms of the indication range that we could target with the GD2 approach. Obviously, there are 2 kind of categories of indications you have, particularly the pediatric indications that tend to be very homogeneously expressed, GD2, which is very attractive, obviously, because it allows you to focus on the additional mechanisms and challenges that the solid tumor may actually provide. Clearly, an interest to actually develop within those settings because it is not only a very high medical need, but it's also enormously informative in terms of the understanding of the behavior of the cells and the ability to impact these -- in these complex diseases. So that's the first part. So the answer is yes, we would like to actually be active in pediatric settings in the GD2 positive. I think, obviously, the larger opportunity, as you indicated, are clearly in indications like small cell lung cancer and this is why we were presenting the data we did today, is that there is a quite significant amount of interest to actually explore the activity of a GD2 targeting approach in these disease settings -- in this disease setting, which is still very challenging to capture with pretty much any treatment modality. And we believe that the modules and the ability to sort of adapt to a very complex and quite diverse [ actually ] environment through the modules that we're providing, I think, puts us in a very good position if we're to sort of track and treat this particular type of disease. So the answer is we want to be active on both sides, and you sort of ask slightly different questions with the respective program. When we look into the prostate cancer field, obviously, there is a lot of activity. There's obviously bispecifics with PSMA that go back to end of use. And in fact, I had the chance to start with my old team, so way back. And we have -- there's a lot of approaches that have been tried, et cetera. Most of them are just single targeting approaches. They have some level of activity. But it's probably fair to say that none of them actually has a profound single-agent activity demonstrated to date. And whether this is in a bite forma, in an ADC format, whether it is a radio conjugate format or whether it's in a CAR T format. And what we do see, clearly, with a disease setting like prostate cancer, it is a challenging disease to actually have a significant impact, particularly in late-stage disease. And so when we look at pretty much all the approaches that are out there, the vast majority of them are single targeting approaches. With one a -- with a specific mechanism of action. I don't think there is, at this point, programs out there that have the -- that look to actually create a product that can adapt to the complexity and challenges of the disease. And I think that is where the approach that we're designing is standing out.

Thank you. This concludes our question-and-answer session. I would now like to turn the call back over to Dr. Christian Itin, our Chairman and Chief Executive Officer, for closing remarks.

All right. Well, thank you very much all for joining today, and I really appreciate you taking the time. We'd like to thank all our presenters for the great presentations. And obviously, we're looking forward to keeping you updated, particularly towards the end of this year, on the progress with our hematological programs, with a particular focus on AUTO1 and AUTO3. And with that, we wish you all the best and looking forward to connecting with you soon. Thank you.

Ladies and gentlemen, this concludes today's conference call. Thank you for participating. You may now disconnect.