Abstract B062: Developing diverse patient-derived xenograft models of common and rare brain metastases to elucidate molecular landscapes and reveal therapeutic opportunities

Abstract Background. Advancing therapeutic approaches to brain metastases (BrMets) is an area of critical need. Preclinical models of BrMets are a rare but much-needed tool to investigate novel therapeutic approaches. We developed a biobank of BrMet patient-derived xenograft (PDX) models established from resected BrMets originating from various solid tumor types. Methods: Resected BrMet tissues were collected. PDX models were established, and ex vivo drug screening was performed using previously published methods (Morikawa et al. Can Res Comm 2023). Molecular profiles of PDXs and matched source tumors were compared, and their correlations with drug response were examined. Results: From Nov 2016 to Sept 2023, 142 surgical cases were collected, of which 126 PDX models were established and maintained growth. Tumor types included common (lung, breast, melanoma) and rare (sarcoma, ovarian, cervical, prostate, renal, and gastrointestinal) tumors. Common mutations across these models included BRCA, ATR, ALK, KMT2C, FAT1, ZFXHX3, MAP3K1, COL6A3, FLT3, MLH1, EGFR, IGFN1, and TP53, though many were not predicted to be pathogenic. In addition, there were less prevalent but potentially targetable alterations, such as PI3K mutations. Copy number variation (CNV) analysis demonstrated a predominance of amplifications over deletions. The observed pathways included those associated with neuronal and structural features such as axonal transport (DNA KEGG database) and extracellular matrix (DNA REACTOME database). The RNA seq analysis revealed clustering mostly based on the primary tumor type. Compared to publicly available metastatic PDX models (NCI database) stratified by primary tumor type, these BrMet models demonstrated differences in the molecular pathway enrichment. PDX models generally exhibited high concordance based on Jaccard Index (JI). The majority of the samples showed JI in the range of 0. 4-0. 6, with melanoma samples demonstrating JI in the lower 0. 2 range. We evaluated the PDX and matched pairs for the selected variants predicted to be pathogenic or possibly functionally impactful. Again, melanoma subtypes showed more divergence, but overall, the majority of the variants were retained in the matched PDX models. Drug sensitivity testing was performed on 13 PDXs using a panel of molecularly targeted agents and chemotherapies. Of 13, nine PDXs had potentially actionable molecular targets. Highly active drugs were identified in these models; however, most drug sensitivities were not predictable based solely on genomic profiles, emphasizing the utility of paired functional characterization. Conclusion: We present a novel biobank of BrMet PDX models. These models provide a valuable resource for probing the biology of BrMet and informing therapeutic strategies. Prospective collection is ongoing to expand the biobank, alongside further studies investigating tumor-microenvironment interactions using immune-competent and organ-on-chip microfluidic blood-brain niche models. AI disclosure: AI was used for language editing only; content was verified by the authors Citation Format: Aki Morikawa, Tusharika Rastogi, Noreen Khan, Peter Ulintz, Derek Nancarrow, Habib Serhan, Xu Cheng, Liwei Bao, Aaron Udager, Matthew Soellner, Jason Heth, Nathan Merrill, Sofia D. Merajver. Developing diverse patient-derived xenograft models of common and rare brain metastases to elucidate molecular landscapes and reveal therapeutic opportunities abstract. In: Proceedings of the AACR Special Conference in Cancer Research: Brain Cancer; 2026 Mar 23-25; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86 (6Suppl): Abstract nr B062.