Molecular Cancer | Molecular Subtyping of Pulmonary Carcinoids Reveals Four Heterogeneous Tumor Classes and Defines the Supra-Carcinoid Biological Signature

This study redefines the molecular classification of pulmonary neuroendocrine tumors through integrative multi-omics analysis, providing key molecular markers and potential therapeutic targets for precision diagnosis and treatment strategies of pulmonary carcinoids, with significant clinical implications especially for identifying high-risk patients.

 

Literature Overview

This article, 'Deep molecular profiling of lung neuroendocrine tumours and supra-carcinoids,' published in the journal Molecular Cancer, systematically investigates the molecular heterogeneity of pulmonary neuroendocrine tumors (NETs) and carcinoids. Based on multi-omics data from 319 fresh-frozen samples, combined with spatial transcriptomics and deep learning-based pathological image analysis, the research team identified four molecular subtypes, among which the newly defined sc-enriched group (i.e., supra-carcinoid) exhibits aggressive clinical behavior and distinct microenvironmental features. This work not only unifies previous molecular classification systems but also provides a clinically translatable molecular-morphological integrated classification framework.

Background Knowledge

Pulmonary carcinoids are neuroendocrine tumors with rapidly increasing incidence but unclear etiology. Their clinical management is limited by the insufficient prognostic discrimination and inter-observer variability of the current WHO grading system (based on mitotic count and necrosis). Driver genes such as TP53 and RB1 are rarely mutated in pulmonary carcinoids, while mutations in MEN1 and BRAF exist but lack clear targeted therapeutic pathways, resulting in extremely limited treatment options. Furthermore, conventional morphology struggles to explain the paradoxical phenomenon where some low-grade tumors exhibit highly aggressive behavior, suggesting the existence of unrecognized high-risk biological subtypes. This study addresses these gaps by integrating genomic, transcriptomic, methylomic, and spatial omics data to systematically dissect the molecular architecture of lung NETs and explore whether supra-carcinoid represents an intermediate state bridging low-grade NETs and high-grade neuroendocrine carcinomas (e.g., SCLC, LCNEC), thereby filling critical knowledge gaps in the heterogeneity and evolutionary trajectory of pulmonary carcinoids.

 

 

Research Methods and Experiments

The research team constructed a large cohort of 319 pulmonary NETs (including paired normal tissues), performing whole-genome sequencing, transcriptome sequencing, DNA methylation arrays, and spatial transcriptomic analyses, integrating the data using MOFA and ParetoTI for multi-omics integration and prototype analysis. To validate the morphological discriminability of molecular subtypes, a deep learning model was applied for unsupervised patch clustering of whole-slide H&E images, with morphological features annotated by six pathologists. Additionally, tumor microenvironment and evolutionary trajectories were dissected using single-cell reference deconvolution, cell interaction analysis, and multi-region sequencing. Key evidence shows that the sc-enriched group not only shares molecular similarities with SCLC but also harbors driver events such as BRAF V600E and TERT amplification. Spatial transcriptomics revealed an ICAM1–ITGB2 interaction axis between LAP-like progenitor cells and macrophages, supporting a tumor-immune microenvironment co-driven oncogenic mechanism.

Key Conclusions and Perspectives

• Integration of multi-omics data defined four molecular subtypes (Ca A1, Ca A2, Ca B, sc-enriched), with the sc-enriched group enriched for supra-carcinoids and associated with poorer overall survival, indicating prognostic value independent of traditional grading and providing a new standard for future clinical stratification.
• sc-enriched tumors exhibit higher mutational burden, structural variations, and genomic instability, along with driver events such as BRAF V600E and TERT amplification, suggesting their genomic evolution resembles that of high-grade neuroendocrine carcinomas, providing a foundation for future targeted therapy exploration.
• Transcriptomic analysis reveals enrichment of myeloid activation signals and M1 macrophages in the sc-enriched group, and spatial analysis confirms co-localization of ICAM1 and ITGB2 between LAP-like cells and macrophages, highlighting tumor-macrophage interactions as potential intervention nodes, offering guidance for future studies on the immune microenvironment.
• The deep learning model accurately identifies molecular subtypes from H&E images, outperforming traditional grading, and identifies morphological features such as spindle cells and fibrotic stroma as associated with specific subtypes, providing a scalable tool for future digital pathology applications.
• Multi-region analysis shows sc-enriched tumors can simultaneously harbor molecular features of Ca A1 or Ca B, and PDTO models demonstrate their potential to transform into LCNEC. Combined with chromothripsis, this supports their role as an intermediate state in the transition from NET to NEC, which is crucial for building future disease evolution models.

Research Significance and Prospects

This study provides clear subtype-specific therapeutic targets for drug development: for example, high expression of DLL3 in Ca A1 suggests potential sensitivity to DLL3-targeted therapies (e.g., Rova-T); the presence of BRAF V600E and TERT amplification in the sc-enriched group supports clinical trials of BRAF/MEK inhibitors. Moreover, the sc-enriched group is significantly enriched for T-cell inflammation signatures, suggesting potential responsiveness to PD-1 inhibitors and offering new directions for immunotherapy strategies.

In clinical monitoring, this molecular classification can be implemented clinically using IHC markers (e.g., ASCL1, HNF1A, OTP), aiding in the identification of high-risk patients and guiding personalized follow-up. Additionally, deployment of deep learning models could enable automated initial screening, improving diagnostic consistency.

For disease modeling, this study supports the development of genetically engineered mouse models carrying BRAF V600E, TERT amplification, and chromothripsis to simulate the evolution of supra-carcinoid. Furthermore, the rapid proliferation and phenotypic plasticity of PDTO models provide an ideal platform for studying resistance mechanisms and drug screening.

 

 

Conclusion

This study systematically reconstructs the molecular landscape of pulmonary carcinoids using deep multi-omics and spatial resolution technologies, establishing supra-carcinoid as an aggressive subtype with distinct biological behaviors and microenvironmental characteristics. Its molecular continuity with high-grade neuroendocrine carcinomas challenges the traditional dichotomy between NET and NEC, proposing a dynamically evolving tumor model. From bench to bedside, this research provides a clinically actionable morphology-molecular integrated classification system, enhancing diagnostic objectivity and reproducibility through deep learning-assisted identification of high-risk cases. More importantly, it reveals targetable nodes such as DLL3, BRAF, and TERT, opening precision therapy avenues for pulmonary carcinoid patients who currently lack effective treatments. In the future, prospective clinical trials based on this classification will validate its value in treatment selection and outcome prediction, potentially reshaping the care paradigm for pulmonary carcinoids and enabling a transition from morphology-based to mechanism-driven precision medicine.

 

Alexandra Sexton-Oates, Émilie Mathian, Noah Candeli, Matthieu Foll, and Lynnette Fernandez-Cuesta. Deep molecular profiling of lung neuroendocrine tumours and supra-carcinoids. Molecular Cancer.