PD model simulations against observed data after a single dose compound 4 (testing slower kinact than the in vitro measurement)

Abstract The high prevalence of ER-positive breast cancer and the emergence of resistance to current endocrine therapies highlight the need for more effective ER degraders. AZD4241 is a novel, potent, orally bioavailable, and selective ER proteolysis-targeting chimera (PROTAC) scheduled to enter clinical evaluation in 2026. We describe preclinical pharmacokinetic/pharmacodynamic (PK/PD) and efficacy modeling that quantifies the relationships among drug exposure, ERα degradation, and antitumor activity, thereby defining the extent of compound plasma exposure and target modulation required to achieve efficacy. These translational analyses are intended to inform dose selection and guide early clinical development of AZD4241.We developed a mechanistic mathematical model to quantify the exposure-target-response profile of AZD4241 in in-vivo patient-derived xenograft (PDX) models harboring either wild-type or mutated ESR1. The PK module characterized plasma concentrations across a range of doses. An indirect-response PK/PD module incorporated AZD4241 mechanism of action, whereby the compound accelerates ERα degradation yielding reductions in total ER protein measured by Western blot. The integrated PK/PD/Efficacy model linked plasma exposure, ERα levels, and tumor growth kinetics in the PDX models. Model parameters were estimated via nonlinear mixed-effects (NLME) modeling using individual longitudinal PK, PD biomarker, and tumor volume data aggregated across multiple studies. The mathematical model captured the dose-dependent reduction of ERα and the associated inhibition of tumor growth observed in the PDX models. The level of ER degradation required to induce tumor regressions on the PDX models was also quantified and will be presented. This work provides quantitative, mechanistic insight into how exposure drives biomarker modulation and antitumor responses, delineating the level of ERα degradation required for robust efficacy in endocrine-sensitive PDX models. The framework supports interpretation of compound-induced PD effects in patients under defined dosing regimens and supplies translational evidence to enable dose selection in early clinical development. Citation Format: Ana Quiroga, Pablo Morentin Gutierrez, Lynet Nyoni, Natalie Cureton, Mandy Lawson, Aaron Smith, Thomas Hayhow, Neil Gibson. Preclinical mechanistic PK/PD/Efficacy modeling for AZD4241, a novel oral estrogen receptor (ER) degrader (PROTAC), to support dose selection during early clinical development [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 5019.

Human PK simulation after a single dose of compounds 1 and 2 over 24h

Epidermal growth factor receptor (EGFR) Exon20 insertions (Exon20Ins) constitute the third most common EGFR activating mutation in non-small cell lung cancer. We developed a semimechanistic pharmacodynamic model for irreversible inhibitors of EGFR Exon20Ins mutations by integrating kinetic data of proprietary compounds with a mechanistic description of EGFR turnover and phosphorylation to investigate the preclinical relationship between phosphorylated EGFR (phosEGFR) reduction and efficacy, and its translation to the clinical setting. In engineered NCI-H2073 cells hosting the Exon20 SVDIns mutation, EGFR turnover was studied via stable isotopic labeling by amino acids in cell culture mass spectrometry and phosEGFR time-course analyzed via ELISA. Kinetic parameters were determined using a biochemical binding assay. These data were integrated into the model to describe phosEGFR inhibition in vitro and in vivo. Tumor volume data from xenograft studies were then used to quantify the relationship between phosEGFR inhibition and antitumor activity. We found that sustained >84% phosEGFR inhibition is required for tumor regression. Clinical phosEGFR simulations were generated for two proprietary inhibitors, providing an early estimation of their active human doses. We also explored clinical phosEGFR reduction induced by the third-generation tyrosine kinase inhibitor osimertinib, suggesting that limited target engagement (TE) may explain modest response achieved in EGFR Exon20Ins at the clinically investigated doses. The developed model is a valuable tool to understand the impact of kinetic characteristics on phosEGFR reduction and related efficacy, select a TE-based criterion for therapeutic dose predictions, and provide interpretation and insights on observed clinical efficacy of irreversible inhibitors in EGFR Exon20Ins.

Son of Sevenless Homologue 1 (SOS1) is a promising oncology target with inhibitors in phase 1/2 clinical studies. A focused HTS triage led to a singular SOS1 series having a pyridyl core. The conformational preference of the diamide pyridyl core was critical to binding potency, leading to pyrazine and pyridyl being the preferred motifs. Application of structure-based design to build into a buried lipophilic pocket led to a significant 50-fold potency enhancement. Strategic fluorination of aryl rings and substituents generated compounds with favorable dipoles, low P-gp and BCRP efflux, and high rat Kpu,u. Multiple analogues were progressed into in vivo PK/PD studies where they were combined with a KRASG12C inhibitor. Combination treated tumors in mice showed deeper, more sustained reductions in DUSP6 mRNA and phosphorylated ERK compared to KRASG12C inhibitor alone. Thus, these novel CNS penetrant SOS1 inhibitors have potential to enhance antitumor responses when combined with RAS or MAPK inhibitors.

Additional in vitro data of compounds 2-4 for model validation

Abstract The androgen receptor (AR) is highly expressed in prostate cancers and is a clinically validated target in oncology. AZD9750 is a novel potent oral selective AR Proteolysis-targeting chimera (PROTAC) with a suitable pharmacological profile to be used in combination with a variety of other therapeutics such as capivasertib, a potent pan-AKT kinase inhibitor with anti-tumor activity in tumors with PIK3CA and PTEN mutations, and saruparib, a PARP1-selective inhibitor especially effective against tumors with mutations in genes like BRCA1 and BRCA2. We present here the preclinical PK/PD/Efficacy modeling work used to understand the anti-tumor mechanism of AZD9750 in combination with AKT and PARP inhibitors. We developed a novel mechanistic mathematical model applied to in vivo preclinical hormone sensitive, ARwt prostate PDX models C901 and MR041. C901 has homologous deletion of BRCA2 and MR041 is PTEN null, making them appropriate candidates for combination with PARP and AKT inhibitors, respectively. The PK module of the model describes the compound exposure in monotherapy and combination. The PD module describes the AR, AKT, GSK3β and S6 total and phosphorylated levels measured by Western Blotting and PARylation levels measured by ELISA. In the efficacy module, the levels of AR, pS6 and PARylation were linked to tumor growth inhibition while pGSK3β and PARylation levels were linked to induction of apoptosis; subsequently, these parameters determine the tumor size. All model parameters were derived from internal studies; some were estimated using Non-Linear Mixed Effect modeling of individual longitudinal PK, PD biomarkers and tumor size data taken from several studies. The model describes well the relationship between plasma concentration of the different compounds and PD biomarkers modulation both in monotherapy and in combination. Furthermore, the mathematical model is capable of explaining the enhanced anti-tumor efficacy in combination as a function of the different biomarkers’ modulation. This study provides quantitative mechanistic insights into the AZD9750 combination with AKT and PARP inhibitors. The study enriches our understanding of biomarkers relevant to AR-PROTACs, PARP inhibitors, and AKT inhibitors, informing the selection of biomarkers for monitoring in clinical trials. Additionally, it quantifies the extent of biomarker modulation required to achieve maximal antitumor activity and supports rational combination strategies, as well as dose and schedule optimization for clinical development. Citation Format: Ana Quiroga, Pablo Morentin Gutierrez, Antonio Ramos-Montoya, Chrysiis Michaloglou, Nuria Galeano-Dalmau, Claire Crafter, Aaron Smith, Jamie Scott, Michael Niedbala. Application of mechanistic preclinical PK/PD/efficacy modeling to support combination strategy for AZD9750, a novel oral androgen receptor degrader (PROTAC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4612.

Optimised transitions from the Waters TQ-XS for compound 1, compound 4, compound 3, compound 2, and clozapine (internal standard) utilising the M+H ion

The androgen receptor (AR) is highly expressed in prostate cancers and is a clinically validated target in oncology. AZD9750 is a novel potent oral selective AR Proteolysis-targeting chimera (PROTAC) due to start clinical testing in 2025. We present here the preclinical PK/PD modeling work used to understand the required target modulation and concentration required to see anti-tumor efficacy and therefore help support the dose selection during the early clinical development of AZD9750. We developed a novel mechanistic mathematical model applied to in vivo preclinical mouse PDX models (MR041 and C901, primary prostatic adenocarcinoma, hormone sensitive and AR-wt) linking the compound pharmacokinetics with the magnitude of target modulation expressed as relative levels of AR measured by Western Blotting. These changes at the biomarker level were subsequently linked to inhibition of tumor cell proliferation resulting in macroscopic dynamic effects on tumor size. All model parameters were derived from internal studies; some were estimated using Non-Linear Mixed Effect modeling of individual longitudinal PK, PD biomarker and tumor size data taken from several studies. The model described well the relationship between plasma concentration of AZD9750 and modulation of AR. In vivo AZD9750 concentration required to see 50% of the maximal target modulation in both animal models was on the region of 0.2-0.7 µM. Population tumor size patterns, for all treatment regimens ranging from 0 to tumour regressions across both tumor models were very well described with the model proposed. 55 - 80% reduction of AR levels was required to induce tumor regression in both models. This study provides quantitative mechanistic insights into the links between compound exposure, biomarker modulation (AR) and anti-tumor responses, supporting our understanding of the required target modulation needed for maximal anti-tumor effects in these two hormone sensitive patient-derived tumor models. This mechanistic understanding is valuable to contextualize compound-induced PD modulation in patients, for given doses and schedules. Pablo Morentin Gutierrez, Ana Quiroga Campano, Antonio Ramos Montoya, Claire Crafter, Chrysiis Michaloglou, Aaron Smith, Nuria Galeano Dalmau, Jamie Scott, Michael Niedbala. Preclinical mechanistic PK/PD-efficacy modeling for AZD9750, a novel oral androgen receptor degrader (PROTAC), to support dose selection during early clinical development [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 4755.

Current clinical therapeutics against non-small cell lung cancer (NSCLC) with epidermal growth factor receptor (EGFR) exon 20 insertion (EGFRExon20Ins) mutations yield limited responses particularly against brain metastases; therefore, there is a need for an effective tyrosine kinase inhibitor (TKI). AZ14289671 is an oral, potent, irreversible, selective, and blood-brain barrier penetrant TKI targeting EGFRExon20Ins mutations while sparing wild-type (WT) EGFR. Preclinical assessments using cell lines, cell line-derived xenograft, and patient-derived xenograft models harboring EGFRExon20Ins demonstrate that AZ14289671 exhibits strong signaling pathway inhibition and highly sustained tumor regression against multiple EGFRExon20Ins, whereas its activity against WT is minimal. Additionally, AZ14289671 can cross the blood-brain barrier. This has the potential to improve outcomes of NSCLC patients with EGFRExon20Ins.

PIK3CA encodes the p110α catalytic subunit of PI3-kinase alpha (PI3Kα) and is the most frequently mutated kinase in human cancer with common mutations occurring in the kinase domain (H1047R) and helical domain (E542K/E545K). The approved PI3Kα inhibitor, alpelisib, shows promise for this targeted class of agents with improvements in progression-free survival in ER+/Her2- breast cancer patients in combination with fulvestrant. However, toxicities attributed to the inhibition of wild-type PI3Kα, such as hyperglycemia, gastrointestinal issues, and skin reactions, lead to sub-optimal target engagement due to requisite dosing modifications. A PI3K mutant inhibitor that spares WT PI3K is predicted to be better tolerated, require fewer dosing modifications, and therefore, have the potential to provide improved clinical benefit. Herein, we present the preclinical characterization of CGT6297, a wild type sparing PI3Kα H1047R inhibitor. Aaron C. Smith, Ben Arwood-Levine, Abiezer Blandon, Alexandra Born, Richard Brizendine, Payal Chatterjee, Mark J. Chicarelli, Michael L. Conner, Brad Fell, Jennifer Fulton, Anna Guarnieri, Hannah Hubert, Ravi Jalluri, Hailey Knox, Keith Koch, Daniel Krischlunas, Vijay Kumar, Sara Kuzbiel, Matt McDonald, Colin McHugh, Brent Mclean, Kelsey W. Nassar, Paul Larsen, Brad Newhouse, Scott Niman, Rob Rieger, John Robinson, Marelí Rodriguez, Leah Salituro, Vinny Scarato, Lee Stunkard, Francis Sullivan, Patrick Sutter, Roy Turton, Rob Van Gulick, Brooklynn Venteicher, Logan Vine, Shannon Winski, Shannon Winski, Hannah Work, Yeyun Zhou. Preclinical characterization of CGT6297, a novel PI3Kα H1047R mutant-selective inhibitor [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3004.

Although combinations of DNA damage response inhibitors (DDRi) and DNA damaging chemotherapy enhance cytotoxicity in cell-based systems, clinical success has been limited by overlapping bone marrow toxicities. Here, we show that a tumor-targeted nanoparticle camptothecin CRLX101, administered concurrently with DDRi, enhances anti-tumour efficacy but also increases bone marrow toxicity in preclinical models. Using rat bone marrow progenitor cells as biomarkers and leveraging differential repair kinetics of CRLX101-induced DNA damage in tumour and bone marrow, we identified a gap schedule of the PARP inhibitor olaparib and CRLX101 that enhanced efficacy over single agents but demonstrated a reduced combination marrow toxicity. A clinical trial has been designed using the gap schedule identified here and represents a template that can be used to successfully deliver DDRi with tumor-targeted chemotherapy in combination. ### Competing Interest Statement MOC, LOC, AW, AS, GH, AL and EC are employees and shareholders of AstraZeneca, while CS, JB, and RO are former employees of AstraZeneca.

Actively targeted nanoparticle systems have the potential to improve delivery to tumors over untargeted systems however the design rules to achieve this have not been fully elucidated. A HER2-targeted polymer drug delivery system composed of a 32-arm star polymer (SD) conjugated with the TOP1 inhibitor molecule SN-38, with a trastuzumab antigen binding fragment (HER2-Fab), has been used to target cancer cells overexpressing this receptor. The HER2-Fab was attached to the SD at two different densities (average of 1 or 3 Fabs per star polymer) and compared to the native star polymer without Fab. In vitro experimentation showed that both the targeted star polymers (HER2-SDs) had better binding and uptake in HER2-positive cell lines (SK-BR3 and HEK293) compared to the non-targeted SD. In vivo biodistribution studies showed enhanced accumulation of HER2-targeted SDs in tumors, but not normal tissues, particularly at the later (96 h post-dose) timepoint. The HER2-SDs demonstrated increased localization with tumor cells rather than in stromal regions, greater penetration into the tumor core and a more homogenous distribution in the tumor section than the untargeted SD. The targeted star polymer conjugated to SN-38 was tested for anti-tumor activity in a HER2-positive gastric cancer xenograft in mice and showed significantly greater efficacy compared to untargeted SDs.

PIK3CA encodes the p110α catalytic subunit of PI3-kinase alpha (PI3Kα) and is the most frequently mutated kinase in human cancer with common mutations occurring in the kinase domain (H1047R) and helical domain (E542K/E545K). The approved PI3Kα inhibitor, alpelisib, shows promise for this targeted class of agents with improvements in progression-free survival in ER+/Her2- breast cancer patients in combination with fulvestrant. However, toxicities attributed to the inhibition of wild-type PI3Kα, such as hyperglycemia, gastrointestinal issues, and skin reactions, lead to sub-optimal target engagement due to requisite dosing modifications. A PI3K mutant inhibitor that spares WT PI3K is predicted to be better tolerated, require fewer dosing modifications, and therefore, have the potential to provide improved clinical benefit. Herein, we present the preclinical in vitro and in vivo activity of a novel, wild type sparing PI3Kα inhibitor series which is potent against the oncogenic H1047R mutation. Citation Format: Aaron C. Smith, Ben Arwood-Levine, Abiezer Blandon, Alexandra Born, Richard Brizendine, Payal Chatterjee, Mark J. Chicarelli, Michael L. Conner, Brad Fell, Jennifer Fulton, Anna Guarnieri, Hannah Hubert, Ravi Jalluri, Hailey J. Knox, Keith Koch, Daniel Krischlunas, Vijay Kumar, Sara Kuzbiel, Colin McHugh, Brent Mclean, Kelsey W. Nassar, Brad Newhouse, Scott Niman, Rob Rieger, John Robinson, Marelí Rodriguez, Leah Salituro, Vincent Scarato, Lee Stunkard, Francis Sullivan, Patrick Sutter, Roy Turton, Robb Van Gulick, Brooklynn Venteicher, Logan E. Vine, Shannon Winski, Yeyun Zhou. Preclinical Characterization of a Novel PI3Kα H1047R Mutant-Selective Inhibitor [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P4-12-19.

PK model fit for AZD3229 dosed orally in the mouse

Pharmacokinetic parameters for standard of care compounds in human and in mouse

Goodness of fit plots of PK/PD model fit for AZD3229

PKPD data utilized to establish the level and duration of target inhibition required for AZD3229