Patient-derived tumor models of tyrosine kinase inhibitor-induced neuroendocrine-transformed lung carcinoma

Background: Lung cancer is the leading cause of cancer-related deaths worldwide, with non-small cell lung cancer (NSCLC) comprising 85% of cases and small cell lung cancer (SCLC) 15%. Tyrosine kinase inhibitors (TKIs) are the frontline therapy for EGFR -mutant and ALK- fusion adenocarcinomas, yet most patients develop resistance. In 5-10% of TKI-resistant cases, adenocarcinomas (ADC) undergo lineage switching to aggressive neuroendocrine (NE) carcinomas, including LCNEC and SCLC; these patients lack effective therapies and have poor clinical outcomes. To better study NE-transformed tumors, we established clinically relevant patient-derived xenograft (PDX) and xenograft-derived organoid (XDO) models to identify novel therapeutic vulnerabilities and elucidate mechanisms underlying NE transformation. Methods: Patient biopsy tissues were subcutaneously engrafted into NOD-SCID mice to generate nine NE-transformed PDX models, and six matched long-term XDOs (>10 passages) were established using organoid culture protocol. PDXs/XDOs were characterized by H&E and immunohistochemistry, whole-exome sequencing, and bulk RNA-sequencing. For in vivo therapeutic profiling, donor PDX tumors were expanded and randomized into four arms: vehicle, Osimertinib (25 mg/kg, QDx5), chemotherapy [Cisplatin (3 mg/kg, QW) + Etoposide (8 mg/kg, TIW)], or combination. For in vitro drug testing, XDOs were dissociated into single cells, seeded into 384-well plates, and treated with Osimertinib, Cisplatin, or Etoposide across a 21-point dilution series (1 nM-10 μM), with viability measured on Day 7 using CellTiter-Glo 3D. Results: Paired PDXs/XDOs retained key genetic drivers, including EGFR, TP53, and RB1 mutations, along with recurrent alterations in the NOTCH signaling pathway. Histologically, the models recapitulated patient tumor phenotypes, including three SCLC-transformed, two LCNEC-transformed (one ADC-like, one NE-like), and one mixed ADC/SCLC case. Notably, the mixed-histology model maintained both ADC and SCLC components in long-term XDO culture, preserving intratumoral heterogeneity. In vivo therapeutic testing revealed minimal response to Osimertinib, chemotherapy, or combination therapy in NE-high tumors, whereas an ADC-like LCNEC PDX exhibited sensitivity to these treatment arms, consistent with clinical observations. XDO drug responses were highly concordant with their matched PDXs, demonstrating resistance to Osimertinib and chemotherapy (IC 50 > 1 μM). Preliminary high-throughput screening of XDOs with a 58-compound epigenetic probe library identified a candidate therapeutic target to be presented. Conclusion: Our NE-transformed models preserve patient tumor molecular features, histopathology, and therapeutic response, providing robust platforms to identify novel therapeutic strategies and dissect mechanisms driving lung NE transformation.