Abstract Introduction: Loss-of-function mutations in the tumour suppressors BRCA1/BRCA2 inactivate the homologous recombination repair (HRR) pathway, driving genomic instability and cancer progression but conferring sensitivity to DNA double-strand break–inducing agents such as platinum salts and poly(ADP-ribose) polymerase inhibitors (PARPi). BRCA1/BRCA2 mutations are frequent in ovarian, breast, pancreatic, and prostate cancers, underpinning approvals of PARPi in these settings. Despite strong activity in HRR-defective tumours, PARPi responses vary and resistance emerges. Attempts to boost efficacy by combining PARPi with other DNA damage response inhibitors have been limited by enhanced myelosuppression, constraining dose intensity and benefit. Results: AZD4956 is a novel, potent, and selective inhibitor of the polymerase activity of DNA polymerase theta (Polθ), a key effector of microhomology-mediated end joining—a repair pathway critical when HRR is compromised. Combining AZD4956 with the PARP1-selective inhibitor saruparib improves efficacy versus either agent alone across HRR-defective cellular backgrounds, with no activity in HRR-proficient settings. Enhanced efficacy correlates with increased genomic instability: in HRR-deficient cells, the combination produces a ∼4-fold rise in chromosomal aberrations. AZD4956 also potentiates other DNA-damaging agents (e.g., cisplatin, TOP1 inhibitors) specifically in HRR-deficient lines.AZD4956 shows single-agent in vitro activity in the low-nanomolar range in cell lines with PALB2, BRCA2, and RAD51C mutations, like PARPi, but no single-agent activity in BRCA1-mutant cells. In vivo, AZD4956 monotherapy yields modest tumour growth inhibition (TGI ∼50–80%) in some BRCA2 and PALB2 mutant models. By contrast, AZD4956 (≥10 mg/kg BID) combined with the maximal efficacious mouse dose of saruparib (1 mg/kg QD) consistently outperforms either monotherapy. Increased efficacy aligns with pharmacodynamic modulation: the combination drives an average ∼2-fold increase in micronuclei in red blood cells versus monotherapy. Combination activity is observed in both BRCA1- and BRCA2-mutant patient-derived xenografts from diverse tissues (breast, prostate) and is restricted to HRR-deficient models where PARPi alone confers some TGI. Notably, maximal combination benefit requires the maximal efficacious saruparib dose. Conclusions: Preclinical pharmacology supports AZD4956 selectivity and its potential to amplify antitumour activity when combined with PARPi in HRR-defective cancers. AZD4956 is being evaluated in PARTHENON, a first-in-human, open-label, multicentre, phase 1/2a study of AZD4956 plus saruparib in patients with HRR-deficient solid tumours. Citation Format: Josep V. Forment, Lee Mulderrig, Christelle de Renty, Harriet Southgate, Gemma Jones, Martina Gesu, Rebecca Sargeant, Daniel Sutton, Lenka Oplustil O'Connor, Susan Critchlow, Sabina Cosulich. AZD4956, a potent and selective inhibitor of DNA polymerase theta, enhances the activity of DNA-damaging agents in HRR defective cellular backgrounds and improves efficacy of the new generation PARP1-selective inhibitor, saruparib [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 230.

Immuno-oncology drug discovery increasingly relies on humanised mouse models of cancer due to limitations of murine surrogate tools and differences between mouse and human immune systems. Graft-versus-Host Disease (GvHD) is a significant complication following xenogeneic transplantation of human immune cells into mice, limiting their lifespan and impacting the utility of these studies. Existing GvHD scoring systems inadequately capture the disease’s complexity, hampering optimal welfare management and clinical progression monitoring. We propose a comprehensive, practical scoring system for monitoring clinical signs of GvHD in humanised mice. It evaluates seven clinical signs reflecting disease complexity, sums the scores, and categorises overall GvHD severity into four stages, each with specific welfare actions. This refined tool reduces animal suffering through early detection and timely interventions, enabling mice to remain on studies where possible to maximise scientific impact. Our scoring system correlates with histological scores of GvHD-induced tissue damage across multiple organs, with liver and kidney histopathology ranking highly, unlike lung pathology. The system is reproducible among independent experimenters and versatile, effectively applied across multiple types of humanised mouse models and strains. It identifies common clinical signs including weight loss, swelling/reddening of extremities, fur condition, and posture changes, aiding users in distinguishing relevant signs. This system refines and standardises welfare decision-making, supporting the responsibility to minimise suffering when working with humanised mice.

The Ataxia telangiectasia and Rad3-related (ATR) inhibitor ceralasertib in combination with the PD-L1 antibody durvalumab demonstrated encouraging clinical benefit in melanoma and lung cancer patients who progressed on immunotherapy. Here we show that modelling of intermittent ceralasertib treatment in mouse tumor models reveals CD8 + T-cell dependent antitumor activity, which is separate from the effects on tumor cells. Ceralasertib suppresses proliferating CD8 + T-cells on treatment which is rapidly reversed off-treatment. Ceralasertib causes up-regulation of type I interferon (IFNI) pathway in cancer patients and in tumor-bearing mice. IFNI is experimentally found to be a major mediator of antitumor activity of ceralasertib in combination with PD-L1 antibody. Improvement of T-cell function after ceralasertib treatment is linked to changes in myeloid cells in the tumor microenvironment. IFNI also promotes anti-proliferative effects of ceralasertib on tumor cells. Here, we report that broad immunomodulatory changes following intermittent ATR inhibition underpins the clinical therapeutic benefit and indicates its wider impact on antitumor immunity.

Camizestrant (AZD9833) is a selective ER degrader and pure antagonist. A, Chemical structure of camizestrant. B, The indicated cell lines were treated with 100 nmol/L of the indicated compound for 48 hours. Levels of ERα were assessed by Western blotting and normalized to an untreated control and fulvestrant. Each point represents an independent experiment. C, MCF7 or CAMA-1 cells were treated with DMSO, 1 nmol/L estradiol, or 100 nmol/L of the indicated compound + 1 nmol/L estradiol for 24 hours. RNA expression was assessed by RNA sequencing. Data represent z-scores of normalized gene expression for genes in an ER activity signature. D, MCF7 and CAMA-1 cells were treated with the indicated concentration of the indicated compound for 7 days. Cell number was estimated with a Sytox Green assay normalized to an untreated control on the day of treatment (0.0) and an untreated control on day 7 after treatment (1.0). Data points represent the mean from three independent experiments performed in triplicate ± SD. E, Ishikawa cells were treated with the indicated concentration of fulvestrant or camizestrant, or 100 nmol/L AZD9496 for 24 hours, and ERα and PgR expressions were determined by Western blot. DMSO, dimethyl sulfoxide; PR, progesterone receptor.

Combining the selective AKT inhibitor, capivasertib, and SERD, fulvestrant improved PFS in a Phase III clinical trial (CAPItello-291), treating HR+ breast cancer patients following aromatase inhibitors, with or without CDK4/6 inhibitors. However, clinical data suggests CDK4/6 treatment may reduce response to subsequent monotherapy endocrine treatment. To support understanding of trials such as CAPItello-291 and gain insight into this emerging population of patients, we explored how CDK4/6 inhibitor treatment influences ER+ breast tumour cell function and response to fulvestrant and capivasertib after CDK4/6 inhibitor treatment. In RB+, RB− T47D and MCF7 palbociclib-resistant cells ER pathway ER and Greb-1 expression were reduced versus naïve cells. PI3K-AKT pathway activation was also modified in RB+ cells, with capivasertib less effective at reducing pS6 in RB+ cells compared to parental cells. Expression profiling of parental versus palbociclib-resistant cells confirmed capivasertib, fulvestrant and the combination differentially impacted gene expression modulation in resistant cells, with different responses seen in T47D and MCF7 cells. Fulvestrant inhibition of ER-dependent genes was reduced. In resistant cells, the combination was less effective at reducing cell cycle genes, but a consistent reduction in cell fraction in S-phase was observed in naïve and resistant cells. Despite modified signalling responses, both RB+ and RB− resistant cells responded to combination treatment despite some reduction in relative efficacy and was effective in vivo in palbociclib-resistant PDX models. Collectively these findings demonstrate that simultaneous inhibition of AKT and ER signalling can be effective in models representing palbociclib resistance despite changes in pathway dependency.

Abstract Background Histopathological endpoints are evolving as a treatment target in Inflammatory Bowel Disease (IBD). Use of histology to screen entrants could add value in IBD clinical trials; for example, by refining eligibility criteria to ensure studies recruit patients with definitive active inflammation at the microscopic level. Several histopathological indices have been developed, but the relative complexity of available scores hinders the development of an AI algorithm without large-scale labour-intensive annotation by a pathologist. We aimed to develop computer vision tools to assist decoding the complex clinical disease features at the histological level for both Crohn’s Disease (CD) and Ulcerative Colitis (UC). This will inform understanding of disease pathology and patient stratification to support clinical trial development strategies. Methods A total of 1397 clinically annotated Haematoxylin & Eosin (H&E) images were included from 418 CD and 218 UC patients enrolled in a multicentred longitudinal Study of a Prospective Adult Research Cohort with IBD (SPARC IBD) obtained from the IBD Plexus program of the Crohn’s & Colitis Foundation. We developed an image quality control (QC) algorithm to automatically identify image and tissue processing/staining artefacts negatively impacting analysis (e.g., out-of-focus, tissue folds, overstained regions) and excluded these regions (Fig. 1). Next, a self-supervised learning (SSL) deep learning computer vison model was developed and trained to predict disease relevant features including disease diagnosis and lesional macroscopic appearance (inflammation, erosions and ulcers). Finally, to better understand the model’s predictions, we generated heatmap overlays on the tissue that show regions which the model considers to be most predictive (Fig. 2) and shared these with pathologists for qualitative evaluation. Results We find that the SSL model performs well on different downstream classification tasks such as UC vs CD (area under curve (AUC) = 0.79) and normal vs lesional tissue (AUC = 0.76). Specialist pathologist collaboration further confirmed that the heatmap overlays identified clinically relevant tissue features, including inflammatory cell infiltrates (Fig. 2). Conclusion These encouraging results support further exploration of this deep-learning algorithm to distinguish disease specific characteristics in this set of images from CD and UC patients. Further work is ongoing to validate the heatmap approach on endoscopic scores, we also plan to validate our model on IBD clinical trial datasets.

Supplementary file containing all supplementary tables, text and figures

Background Keynote-942 (AACR 2023; ASCO 2023) represents the first clinical success of a personalized cancer vaccine. mRNA-4157 is a personalized neoantigen vaccine (Moderna) that, when combined with pembrolizumab, demonstrated prolongation of RFS and DMFS when given as adjuvant therapy for high-risk melanoma following complete resection. These findings demonstrate that the immune system can be successfully primed to target cancer neoantigens for clinical benefit; however, individualized vaccines are costly in time and manufacturing, limiting generalizability. Methods An alternative approach is to focus on shared antigens that arise across a large frequency of patients, enabling an 'off-the-shelf' targeted immunotherapy. Success factors for such an approach include identification of shared (neo)antigens that are: (1) present in large numbers of patients, (2) specific to tumors, and (3) under evolutionary pressure to be retained. Acquired resistance mutations (ARM) fulfill these criteria as they: (1) arise in the majority of patients undergoing targeted therapy, (2) are specific or over-represented in tumors, (3) are retained as a result of the selective pressure from the targeted therapeutic, (4) have clear prognostic value as they drive treatment resistance, and (5) have not been clinically responsive to existing checkpoint inhibitors. Thus, ARMs are ideal as targets for innovative approaches to cancer immunotherapy. Results Replicate Bioscience has developed a precision immunotherapy (PIO), RBI-1000, targeting ARM in ER+ breast cancer. RBI-1000 is a self-replicating RNA (16,000 bases on a novel alphaviral vector encoded in a lipid nanoparticle) encoding multiple on-target and bypass mutations that arise as patients are on 1L endocrine therapy (e.g., ESR1m+). Immune responses, inclusive of both antibodies and T cells, elicited by RBI-1000 leads to control and elimination of tumor cells expressing ARM in preclinical models. Coupling targeted therapy SOC and PIO (i.e. estrogen blockade and RBI-1000) generates a synthetic immune lethal state for the tumor: if the tumor retains endogenous ESR1, it is subject to the targeted therapy, while if it develops an ARM, it is now eliminated by RBI-1000. RBI-3000, our EGFRm PIO and second oncology therapeutic, targets the known resistance mutations, and primary mutations that are less responsive to tyrosine kinase inhibitor (TKI) therapy and will be administered in combination with SOC TKIs. The coupling of targeted and PIO therapies is anticipated to better address clinical need in the metastatic setting, with chemotherapy replacing surgical resection to address tumor bulk. Conclusions PIO is a novel approach to cancer immunotherapy that is widely applicable to any cancer with characterized ARM.

Camizestrant has superior in vivo activity to fulvestrant in ESR1wt and ESR1m PDX models (2). A, Characteristics of ER+ breast cancer models used. B, Change in ER pathway gene activation after treatment, expressed as change in ER pathway gene score and in cell-cycle G1–S checkpoint genes. See Supplementary Methods for details. Statistical analysis comparing fulvestrant and camizestrant was done using one-way analysis of covariance (n ≥ 4 animals per group). Models shown in bold (x-axis) are fulvestrant sensitive; those in regular type are fulvestrant resistant. ***, P < 0.001; ****, P < 0.0001. Amp, amplification; CCND1, cyclin D1; Del, deletion; MET, metastasis; Mut, mutation; PIK3CA, phosphatidylinositol 3-kinase subunit α; PR, progesterone receptor; PRIM, primary; RESIST, resistant; RB1, retinoblastoma gene; SENS, sensitive. Patient treatment reported: Be, bevacizumab; Ch, chemotherapy; E, exemestane; Ev, everolimus; F, fulvestrant; I, investigational; L, letrozole, T, tamoxifen; X, radiotherapy.

AZD9592 is a bispecific antibody drug conjugate (ADC) designed to deliver a topoisomerase 1 inhibitor (TOP1i) cytotoxic payload (AZ14170133) to tumor cells. AZD9592 selectively binds to epidermal growth factor receptor (EGFR) and c-MET, two cell surface receptors highly expressed in solid tumors including non-small-cell lung cancer (NSCLC) and head and neck squamous cell carcinoma (HNSCC). Here we evaluate the pharmacodynamic activity of the TOP1i payload delivery by AZD9592 in an NSCLC-derived xenograft model using immunohistochemistry (IHC) approaches. Treatment-induced DNA double-strand breaks (DSB) and apoptotic cell death were measured using γH2AX, phospho-RAD-50 (pRAD50), and cleaved-caspase-3 (CC3) across increasing exposure to AZD9592. Furthermore, we report in vivo antitumor efficacy of AZD9592 in a panel of NSCLC and HNSCC patient-derived xenograft (PDX) models that were characterized for somatic tumor alterations, including oncogenic driver mutations in EGFR, tumor cell expression of EGFR and c-MET by IHC and deep-learning based image analysis, and targeted proteomics by mass spectrometry. Results demonstrate dose-dependent increases in pRAD50 and γH2AX upon treatment with AZD9592, signifying induction of DNA damage. Increased CC3 and reduced tumor volume (TV) in all treatment groups compared with control groups supports that AZD9592 induces tumor cell death due to formation of DNA DSB. In PDX experiments, tumor growth inhibition (TGI), defined as ≥30% reduction in TV from baseline after a single dose of AZD9592 8 mg/kg, was observed in 73% (16/22) of EGFR mutant NSCLC models. The models evaluated included tumors with or without prior exposure to EGFR tyrosine kinase inhibitors, and harboring diverse mutational profiles and heterogeneous expression levels of EGFR and c-MET. In EGFR wildtype NSCLC and HNSCC PDX, TGI was observed in 60% (12/20) and 44% (4/9) of models, respectively. IHC demonstrated an association of target expression and response to treatment, suggesting a potential predictive feature of response in tumors with elevated antigen expression. Targeted proteomics demonstrated an association between the expression of SLFN11, a known TOP1i sensitivity marker, and treatment response. Collectively, these results support the hypothesized mechanism of action of AZD9592: TOP1i induced tumor cell death due to formation of DNA DSB, and suggest opportunities in the treatment of tumors with a range of molecular features. AZD9592 is currently in a Phase 1 clinical trial in advanced solid malignancies. Citation Format: Lara McGrath, Ying Zheng, Simon Christ, Christian C. Sachs, Sihem Khelifa, Claudia Windmüller, Steve Sweet, Yeoun Jin Kim, Daniel Sutton, Michal Sulikowski, Arthur Lewis, Ivan Inigo, Nicolas Floch, Edward Rosfjord, Fernanda Arnaldez, Frank Comer. Evaluation of the relationship between target expression and in vivo anti-tumor efficacy of AZD9592, an EGFR/c-MET targeted bispecific antibody drug conjugate. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5737.

Binding and activity of camizestrant in clinically relevant ERα mutations. A, The pIC50 value of fulvestrant and camizestrant to displace a fluorescent ER ligand from wild-type, D538G, Y537N, E380Q, Y537C, S463P, or Y537S mutant purified ERα ligand-binding domain. Points represent independent experiments. B, MCF7 cells expressing WT or Y537S ERα were grown for 7 days in 5% FBS. Growth inhibition was estimated with a Sytox Green assay normalized to an untreated control on day 0 (0%) and an untreated control on day 7 of treatment (100%). Data points represent the mean from two independent experiments carried out in duplicate. Fulvestrant and camizestrant inhibited the proliferation of both WT and Y537S ERα-expressing MCF7 cells in a concentration-dependent manner. The table shows pIC50 values from independent experiments. C, MCF7 cells expressing WT or Y537S ERα were treated with the indicated concentration of fulvestrant or camizestrant for 72 hours, and ERα was determined by Western blot. Fulvestrant and camizestrant showed concentration-dependent inhibition of PgR expression (normalized to an untreated control) in MCF7 cells expressing both WT and Y537S ERα. D, Camizestrant dose–response in the long-term estrogen-deprived ESR1wt PDX model, HBXF079-LTED. Statistical analysis was performed by one-tailed, unequal variance t test versus log (change in tumor volume) compared with vehicle control at the final day of treatment. E, ER degradation measured by Western blot from tumors taken at the end of the efficacy dosing period. F, In the ESR1m D538G PDX CTC-174 model, camizestrant demonstrated antitumor activity in a dose-dependent manner, with maximal antitumor activity at 10 mg/kg. Efficacy correlated with ER degradation measured by Western blot from tumors taken at the end of the efficacy dosing period. Statistical analyses were performed by one-tailed, unequal variance t test versus log (change in tumor volume) compared with vehicle control at the final day of treatment. G, ER degradation measured by Western blot from tumors taken at the end of the efficacy dosing period. NS, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. pIC50, negative log of the IC50 (half-maximal inhibitory concentration) value when converted to mol/L.

Crohn's Disease (CD) and Ulcerative Colitis (UC) are the two main Inflammatory Bowel Disease (IBD) types. We developed deep learning models to identify histological disease features for both CD and UC using only endoscopic labels. We explored fine-tuning and end-to-end training of two state-of-the-art self-supervised models for predicting three different endoscopic categories (i) CD vs UC (AUC=0.87), (ii) normal vs lesional (AUC=0.81), (iii) low vs high disease severity score (AUC=0.80). We produced visual attention maps to interpret what the models learned and validated them with the support of a pathologist, where we observed a strong association between the models' predictions and histopathological inflammatory features of the disease. Additionally, we identified several cases where the model incorrectly predicted normal samples as lesional but were correct on the microscopic level when reviewed by the pathologist. This tendency of histological presentation to be more severe than endoscopic presentation was previously published in the literature. In parallel, we utilised a model trained on the Colon Nuclei Identification and Counting (CoNIC) dataset to predict and explore 6 cell populations. We observed correlation between areas enriched with the predicted immune cells in biopsies and the pathologist's feedback on the attention maps. Finally, we identified several cell level features indicative of disease severity in CD and UC. These models can enhance our understanding about the pathology behind IBD and can shape our strategies for patient stratification in clinical trials.

Enhanced efficacy of camizestrant in combination with PI3K/AKT/mTOR inhibitors as doublets in CDK4/6i-resistant models (2). A, 28-, 35- or 42-day efficacy studies using several ER+ breast cancer PDX harboring/not harboring alterations in PIK3CA/AKT/PTEN. Dark blue, mutations; light blue, deletions; orange, fusions. The rate of growth for each animal is estimated on the basis of fitting each tumor's growth curve to an exponential model: log10(tumor volume) = a + b·time + error, where a and b correspond to the log initial volume and growth rate, respectively. The model assumes that the error terms are normally distributed. Tumor volumes less than 15 mm3 were replaced with a minimum value of 15 mm3. This growth rate summary metric was then used for statistical analysis to compare treatments with a user-specified reference group. Tumor growth inhibition was used to plot a heat map. Designed dosing: oral palbociclib 50 mg/kg daily, subcutaneous fulvestrant 5 mg weekly, oral camizestrant 10 mg/kg daily, oral capivasertib 130 mg/kg BID 4 days on/3 days off. Statistical analysis was performed by one-tailed, unequal variance t test versus log (change in tumor volume) at the final day of treatment. B, 28-, 35- or 42-day efficacy studies used in A; relative tumor volume plots displaying arms: control, standard-of-care hormone therapy + CDK4/6 inhibitor (fulvestrant + palbociclib), or triplet combination of hormone therapy + CDK4/6 inhibitor + AKTi (camizestrant + palbociclib + capivasertib). Designed dosing: oral palbociclib 50 mg/kg daily, subcutaneous fulvestrant 5 mg weekly, oral camizestrant 10 mg/kg daily, oral capivasertib 130 mg/kg BID 4 days on/3 days off. Statistical analysis was performed by one-tailed, unequal variance t test versus log (change in tumor volume) at the final day of treatment. C, Camizestrant fits centrally in the overall landscape of breast cancer as a backbone endocrine therapy. Estrogens (e.g., E2) bind to ERα, leading to its dimerization and translocation to the nucleus, where ERα dimers bind to coactivators to form transcriptionally active ERα complexes. Activated complexes regulate gene transcription in the nucleus or activate kinases in the cytoplasm to drive cell proliferation. Mutations in the ligand-binding domain of ESR1 drive resistance in advanced ER+ breast cancer and act independently of estrogens to activate transcription. Camizestrant is a next-generation SERD for the treatment of ER+ breast cancer, acting as a pure ER antagonist and selective ERα degrader. Camizestrant's mechanism of action stops the transcription of ER target genes in wild-type (blue) and mutant (green) ERα, impairing tumor cell proliferation. These properties position camizestrant as a central endocrine therapy partner along with CDK4/6 inhibitors (palbociclib and abemaciclib) in ER+ breast cancer. Other signaling pathways are essential to ER+ breast cancer proliferation and survival, and contribute to mechanisms of endocrine therapy resistance, including CDK4/6 and PI3K/AKT/mTOR pathways. Inhibitors of these signaling axes are currently approved targeted therapies (everolimus and alpelisib) or under investigation (e.g., capivasertib). *, P < 0.05; **, P < 0.005; ***, P < 0.0005. CAMI, camizestrant; CAPI, capivasertib; CDK, cyclin-dependent kinase; CoA, cytochrome C oxidase assembly; Del, deletion; E2, estradiol; E2F, E2F transcription factor; ERE, estrogen response element; FULV, fulvestrant; m, mutated; MET, metastatic; PALBO, palbociclib; PRIM, primary; RB1, retinoblastoma gene; TGI, tumor growth inhibition.

Abstract Oral selective estrogen receptor degraders (SERD) could become the backbone of endocrine therapy (ET) for estrogen receptor–positive (ER+) breast cancer, as they achieve greater inhibition of ER-driven cancers than current ETs and overcome key resistance mechanisms. In this study, we evaluated the preclinical pharmacology and efficacy of the next-generation oral SERD camizestrant (AZD9833) and assessed ER–co-targeting strategies by combining camizestrant with CDK4/6 inhibitors (CDK4/6i) and PI3K/AKT/mTOR-targeted therapy in models of progression on CDK4/6i and/or ET. Camizestrant demonstrated robust and selective ER degradation, modulated ER-regulated gene expression, and induced complete ER antagonism and significant antiproliferation activity in ESR1 wild-type (ESR1wt) and mutant (ESR1m) breast cancer cell lines and patient-derived xenograft (PDX) models. Camizestrant also delivered strong antitumor activity in fulvestrant-resistant ESR1wt and ESR1m PDX models. Evaluation of camizestrant in combination with CDK4/6i (palbociclib or abemaciclib) in CDK4/6-naive and -resistant models, as well as in combination with PI3Kαi (alpelisib), mTORi (everolimus), or AKTi (capivasertib), indicated that camizestrant was active with CDK4/6i or PI3K/AKT/mTORi and that antitumor activity was further increased by the triple combination. The response was observed independently of PI3K pathway mutation status. Overall, camizestrant shows strong and broad antitumor activity in ER+ breast cancer as a monotherapy and when combined with CDK4/6i and PI3K/AKT/mTORi. Significance: Camizestrant, a next-generation oral SERD, shows promise in preclinical models of ER+ breast cancer alone and in combination with CDK4/6 and PI3K/AKT/mTOR inhibitors to address endocrine resistance, a current barrier to treatment.

Genetically engineered medicines such as chimeric antigen receptor (CAR) T cells have great potential to be the next pillar of medical therapy beyond chemo- and traditional biologic therapies. To develop genetic medicines, new methods to understand their pharmacokinetics (PK) in humans are crucial. It is not feasible to perform traditional PK analysis for “living drugs”, because the genes themselves (in the form of DNA or RNA), are not typically responsible for the therapeutic effect. Rather, the protein products of the genes or the cells harboring the engineered genes are the actuators, and thus cannot be measured using standard HPLC or ligand binding immunoassays for PK analysis. We used a positron emission tomography (PET) reporter gene or “imaging tag” based on the intracellular bacterial enzyme dihydrofolate reductase (eDHFR) that can be paired with radiolabeled versions of trimethoprim (TMP). In this work, we evaluate the potential for immunogenicity using primary human cells and assays geared to assess low affinity and rare T cell clones that may react to eDHFR. We used overlapping pools of 15-mer eDHFR peptides and found that across 9 patients, there was little reactivity compared to EBV and CMV peptide controls. Further, the relative strength of reactivity to the eDHFR peptides was less than that of the viral peptides. Next, we showed that eDHFR iTag harboring CAR T cells were functionally comparable to unlabeled CAR T cells in vitro, and demonstrated strong, selective [18F]-TMP uptake in the eDHFR-expressing CAR T cells. Finally, using a glypican 3 (GPC3) CAR T rodent model, we performed a feasibility study to non-invasively track proliferation in antigen-harboring xenograft tumors over time with ex vivo correlation to anti-CD3 immunohistochemistry. These data demonstrate the potential for non-invasive monitoring of CAR T cells using PET imaging and translational applicability of DHFR/TMP radiotracers. Citation Format: Mark A Sellmyer, Iris K Lee, Kyle Kuszpit, Jyoti Roy, Alex Alfaro, Virginie Ory, Lily Cheng, Daniel Sutton, Emily Bosco, Christine Fazenbaker, Shabazz Novarra, Ryan Gilbreth, Nick Tschernia, Deborah Berry, Xiaoru Chen, Yuling Wu, Ryan Wong. Evaluation of eDHFR/iTag PET reporter gene immunogenicity and application in GPC3 CAR T cells [abstract]. In: Proceedings of the AACR Special Conference: Tumor Immunology and Immunotherapy; 2022 Oct 21-24; Boston, MA. Philadelphia (PA): AACR; Cancer Immunol Res 2022;10(12 Suppl):Abstract nr A37.

1 Imaging and Data Analytics, Clinical Pharmacology & Safety Sciences, R&D, AstraZeneca, Cambridge, UK 2 Department of Chemistry, Technical University of Munich, Garching, Germany 3 Bioscience, Discovery, Oncology R&D, AstraZeneca, Cambridge, UK 4 Institute of Medical Bioinformatics and Biostatistics, University of Marburg, Germany 5 Translational Pathology & Biomarker Analysis, Translational Medicine, R&D Oncology, AstraZeneca, Gaithersburg, MD, USA 6 Antibody Discovery and Protein Engineering (ADPE), R&D, AstraZeneca, Cambridge, UK 7 DMPK, Oncology R&D, AstraZeneca, Cambridge, UK 8 National Centre of Excellence in Mass Spectrometry Imaging (NiCE-MSI), National Physical Laboratory, Teddington, UK 9 Institute of Infection, Immunity and Inflammation, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, UK 10 Advanced Drug Delivery, Pharmaceutical Sciences, R&D, AstraZeneca, Macclesfield UK

Antibody-Drug-Conjugates (ADCs) are biopharmaceutical drugs designed for targeted tumor therapy, meant to improve therapeutic index by restricting drug delivery to tumor cells that express the target antigen. ADCs bind to the target molecule on the cell membrane, which triggers internalization, linker cleavage, and ultimately drug release inside the target cell. Prospective patient selection can be done by quantifying the level of target expression in the tumor using immuno-histochemistry (IHC). However, this process typically involves pathologists and is time consuming, expensive, and prone to human bias. We have developed a supervised deep learning algorithm that segments IHC images of invasive tumor epithelium into individual epithelial cells and their membrane, cytoplasm and nucleus with high accuracy. On unseen test data its performance for epithelial cell detection and segmentation is comparable to the inter-pathologist consensus. With our algorithm, we can describe the target molecule distribution of individual cells in a fully quantitative fashion after using standard IHC methods: we call our approach Quantitative Continuous Score (QCS). We applied QCS to interrogate the mechanism of action of AZD8205, a B7-H4 directed ADC incorporating a novel topoisomerase I linker-warhead. Pharmacodynamic effects were evaluated in vivo, using a human tumor xenograft mouse model and the cell line HT29-huB7-H4 Clone 26, engineered to express human B7-H4. After tumors grew in volume to approximately 250 to 300 mm3, animals were randomized and each mouse received an IV injection of either AZD8205 (1.25, 3.5, or 7 mg/kg) or control articles. Tumors were collected at designated timepoints, fixed in 10% neutral buffered formalin and subsequently embedded into paraffin blocks. IHC and QCS were then used to examine human IgG, γH2AX foci, cleaved caspase-3, and epithelial cell density in tumor samples over time. Using our novel approach we could quantitatively measure the level of AZD8205 bound to tumor cells, with the highest level of ADC on the cell membrane detected at 24-48 hrs. Increased dose levels accelerated the binding kinetics of the drug and led to to a 4- and 3-fold excess of γH2AX and CC-3 respectively, as well as more cells being killed, with up to 2/3 of all epithelial cells dead at the highest dose studied. In summary, we here set the basis for future mechanistic investigation of model systems using computational pathology to improve our understanding of ADC effects. Computational pathology has the potential to determine molecule abundance quantitatively, increase throughput and avoid human bias. Our data implies QCS has the potential to identify patients who may respond to AZD8205, which we will interrogate further and integrate into future clinical studies. Citation Format: Philipp Wortmann, Tze Heng Tan, Susanne Haneder, Andrea Ennio Storti, Ansh Kapil, Jon Chesebrough, Daniel Sutton, Michal Sulikowski, Arthur Lewis, Sofia Koch, Steve Sweet, Zifeng Song, David Chain, Yeoun Jin Kim, Nadia Luheshi, Krista Kinneer, Zachary A. Cooper, Marlon Rebelatto, Günter Schmidt, Hadassah Sade, J. Carl Barrett. Development and implementation of image analysis-based Quantitative Continuous Score (QCS) for B7-H4 IHC to understand AZD8205 pharmacodynamics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 452.

Gaining insight into the heterogeneity of nanoparticle drug distribution within tumors would improve both design and clinical translation of nanomedicines. There is little data showing the spatio-temporal behavior of nanomedicines in tissues as current methods are not able to provide a comprehensive view of the nanomedicine distribution, released drug or its effects in the context of a complex tissue microenvironment. Methods: A new experimental approach which integrates the molecular imaging and bioanalytical technologies MSI and IMC was developed to determine the biodistribution of total drug and drug metabolite delivered via PLA-PEG nanoparticles and to overlay this with imaging of the nanomedicine in the context of detailed tumor microenvironment markers. This was used to assess the nanomedicine AZD2811 in animals bearing three different pre-clinical PDX tumors. Results: This new approach delivered new insights into the nanoparticle/drug biodistribution. Mass spectrometry imaging was able to differentiate the tumor distribution of co-dosed deuterated non-nanoparticle-formulated free drug alongside the nanoparticle-formulated drug by directly visualizing both delivery approaches within the same animal or tissue. While the IV delivered free drug was uniformly distributed, the nanomedicine delivered drug was heterogeneous. By staining for multiple biomarkers of the tumor microenvironment on the same tumor sections using imaging mass cytometry, co-registering and integrating data from both imaging modalities it was possible to determine the features in regions with highest nanomedicine distribution. Nanomedicine delivered drug was associated with regions higher in macrophages, as well as more stromal regions of the tumor. Such a comparison of complementary molecular data allows delineation of drug abundance in individual cell types and in stroma. Conclusions: This multi-modal imaging solution offers researchers a better understanding of drug and nanocarrier distribution in complex tissues and enables data-driven drug carrier design.

Gemcitabine (dFdC) is a common treatment for pancreatic cancer; however, it is thought that treatment may fail because tumor stroma prevents drug distribution to tumor cells. Gemcitabine is a pro-drug with active metabolites generated intracellularly; therefore, visualizing the distribution of parent drug as well as its metabolites is important. A multimodal imaging approach was developed using spatially coregistered mass spectrometry imaging (MSI), imaging mass cytometry (IMC), multiplex immunofluorescence microscopy (mIF), and hematoxylin and eosin (H&E) staining to assess the local distribution and metabolism of gemcitabine in tumors from a genetically engineered mouse model of pancreatic cancer (KPC) allowing for comparisons between effects in the tumor tissue and its microenvironment. Mass spectrometry imaging (MSI) enabled the visualization of the distribution of gemcitabine (100 mg/kg), its phosphorylated metabolites dFdCMP, dFdCDP and dFdCTP, and the inactive metabolite dFdU. Distribution was compared to small-molecule ATR inhibitor AZD6738 (25 mg/kg), which was codosed. Gemcitabine metabolites showed heterogeneous distribution within the tumor, which was different from the parent compound. The highest abundance of dFdCMP, dFdCDP, and dFdCTP correlated with distribution of endogenous AMP, ADP, and ATP in viable tumor cell regions, showing that gemcitabine active metabolites are reaching the tumor cell compartment, while AZD6738 was located to nonviable tumor regions. The method revealed that the generation of active, phosphorylated dFdC metabolites as well as treatment-induced DNA damage primarily correlated with sites of high proliferation in KPC PDAC tumor tissue, rather than sites of high parent drug abundance.

Imaging mass cytometry (IMC) offers the opportunity to image metal- and heavy halogen-containing xenobiotics in a highly multiplexed experiment with other immunochemistry-based reagents to distinguish uptake into different tissue structures or cell types. However, in practice, many xenobiotics are not amenable to this analysis, as any compound which is not bound to the tissue matrix will delocalize during aqueous sample-processing steps required for IMC analysis. Here, we present a strategy to perform IMC experiments on a water-soluble polysarcosine-modified dendrimer drug-delivery system (S-Dends). This strategy involves two consecutive imaging acquisitions on the same tissue section using the same instrumental platform, an initial laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MSI) experiment followed by tissue staining and a standard IMC experiment. We demonstrated that settings can be found for the initial ablation step that leave sufficient residual tissue for subsequent antibody staining and visualization. This workflow results in lateral resolution for the S-Dends of 2 μm followed by imaging of metal-tagged antibodies at 1 μm.