大细胞神经内分泌癌对Ras和ASCL1致癌通路的双重依赖

Background: Large cell neuroendocrine carcinomas (LCNECs) harbor frequent activating alterations in Ras and Raf family members, which represent targetable oncogenes in non-small cell lung cancers (NSCLCs) lacking neuroendocrine (NE) gene expression. LCNECs also overexpress neurodevelopmental transcription factors such as Achaete-scute homolog 1 (ASCL1), which act as oncogenes in small cell lung cancers (SCLCs) and drive NE gene expression. We investigate whether these pathways represent dependencies in LCNEC that could be therapeutically targeted. Methods: Most well-characterized lung cancer cell lines predate the introduction of LCNEC into the World Health Organization (WHO) Classification of Tumors in 2015. To address the scarcity of bona fide LCNEC models, candidate lung cancer cell lines were identified by genomic and transcriptomic features and credentialed rigorously through blinded review of cell line-derived xenografts by a multi-institutional panel of thoracic pathologists. In validated LCNEC lines, Ras pathway dependence was tested by comparing sensitivity to pharmacologic inhibitors with non-NE NSCLC lines harboring the same mutations, using rescue alleles to rule out off-target cytotoxicity. In parallel, ASCL1 dependence was assessed by survival after CRISPR-Cas9 gene deletion, and the therapeutic potential of targeting ASCL1-driven NE gene expression was explored with agents against Delta-like Ligand 3 (DLL3), a transcriptional target of ASCL1. Results: LCNEC cell lines with Ras or Raf activation were sensitive to oncogene-targeted inhibitors. Among 20 lung cancer lines with oncogenic Ras alterations, efficacy of the pan-Ras inhibitor RMC-6236 did not differ between LCNEC (8) and non-NE NSCLC (12) models. In contrast, lines with alterations that bypass Ras, such as activating BRAF mutations, were comparably resistant to RMC-6236 regardless of NE expression. The first LCNEC lines to be re-credentialed (HCC1833, NCI-H1755, and NCI-H1385) expressed high ASCL1 and harbored activating Ras or Raf alterations. Each showed sensitivity to oncogene-targeted inhibitors comparable to non-NE models with the same alterations, and this sensitivity was rescued by on-target or bypass mutations. Proliferation of HCC1833 was greatly reduced by CRISPR-mediated ASCL1 loss, and its sensitivity to tarlatamab, a bispecific T-cell engager (BiTE) against DLL3, was comparable to the most responsive SCLC models. Conclusions: Human LCNEC models demonstrate frequent dependence on both the Ras-activated kinase cascade and neurodevelopmental transcription factors such as ASCL1. This is unexpected, as forced Ras expression inhibits SCLC proliferation, and forced ASCL1 expression does the same in lung adenocarcinoma. Dual dependence on the Ras-activated kinase cascade and ASCL1 may be a defining feature of LCNEC, and we are now exploring therapeutic strategies to exploit this vulnerability.