RESUMEN
The determination of molecular features that mediate clinically aggressive phenotypes in prostate cancer remains a major biological and clinical challenge1,2. Recent advances in interpretability of machine learning models as applied to biomedical problems may enable discovery and prediction in clinical cancer genomics3-5. Here we developed P-NET-a biologically informed deep learning model-to stratify patients with prostate cancer by treatment-resistance state and evaluate molecular drivers of treatment resistance for therapeutic targeting through complete model interpretability. We demonstrate that P-NET can predict cancer state using molecular data with a performance that is superior to other modelling approaches. Moreover, the biological interpretability within P-NET revealed established and novel molecularly altered candidates, such as MDM4 and FGFR1, which were implicated in predicting advanced disease and validated in vitro. Broadly, biologically informed fully interpretable neural networks enable preclinical discovery and clinical prediction in prostate cancer and may have general applicability across cancer types.
Asunto(s)
Aprendizaje Profundo , Neoplasias de la Próstata/diagnóstico , Neoplasias de la Próstata/tratamiento farmacológico , Proteínas de Ciclo Celular/genética , Resistencia a Antineoplásicos/efectos de los fármacos , Resistencia a Antineoplásicos/genética , Humanos , Masculino , Neoplasias de la Próstata/genética , Proteínas Proto-Oncogénicas/genética , Receptor Tipo 1 de Factor de Crecimiento de Fibroblastos/genética , Receptores Androgénicos/genética , Reproducibilidad de los Resultados , Proteína p53 Supresora de Tumor/genéticaRESUMEN
Chimeric antigen receptor (CAR) T-cell therapy can lead to dramatic clinical responses in B-cell malignancies. However, early clinical trials with CAR T-cell therapy in non-B-cell malignancies have been disappointing to date, suggesting that tumor-intrinsic features contribute to resistance. To investigate tumor-intrinsic modes of resistance, we performed genome scale CRISPR-Cas9 screens in mesothelin (MSLN)-expressing pancreatic cancer cells. Co-culture with MSLN-targeting CAR T cells identified both antigen-dependent and antigen-independent modes of resistance. In particular, loss of the majority of the genes involved in the pathway responsible for GPI-anchor biosynthesis and attachment abrogated the ability of CAR T cells to target pancreatic cancer cells, suggesting that disruption of this pathway may permit MSLN CAR T-cell evasion in the clinic. Antigen-independent mediators of CAR T-cell response included members of the death receptor pathway as well as genes that regulate tumor transcriptional responses, including TFAP4 and INTS12. TFAP4-mediated CAR T resistance depended on the NFκB transcription factor p65, indicating that tumor resistance to CAR T-cell therapy likely involves alterations in tumor-intrinsic states. Overall, this study uncovers multiple antigen-dependent and -independent mechanisms of CAR T-cell evasion by pancreatic cancer, paving the way for overcoming resistance in this disease that is notoriously refractory to immunotherapy. SIGNIFICANCE: The identification and validation of key determinants of CAR T-cell response in pancreatic cancer provide insights into the landscape of tumor cell intrinsic resistance mechanisms and into approaches to improve therapeutic efficacy.
Asunto(s)
Inmunoterapia Adoptiva , Neoplasias Pancreáticas , Receptores Quiméricos de Antígenos , Humanos , Línea Celular Tumoral , Tratamiento Basado en Trasplante de Células y Tejidos , Neoplasias Pancreáticas/patología , Neoplasias Pancreáticas/terapia , Receptores de Antígenos de Linfocitos T , Receptores Quiméricos de Antígenos/genética , Receptores Quiméricos de Antígenos/uso terapéutico , Neoplasias PancreáticasRESUMEN
A powerful feature of adaptive memory is its inherent flexibility. Alcohol and other addictive substances can remold neural circuits important for memory to reduce this flexibility. However, the mechanism through which pertinent circuits are selected and shaped remains unclear. We show that circuits required for alcohol-associated preference shift from population level dopaminergic activation to select dopamine neurons that predict behavioral choice in Drosophila melanogaster. During memory expression, subsets of dopamine neurons directly and indirectly modulate the activity of interconnected glutamatergic and cholinergic mushroom body output neurons (MBON). Transsynaptic tracing of neurons important for memory expression revealed a convergent center of memory consolidation within the mushroom body (MB) implicated in arousal, and a structure outside the MB implicated in integration of naïve and learned responses. These findings provide a circuit framework through which dopamine neuronal activation shifts from reward delivery to cue onset, and provide insight into the maladaptive nature of memory.