Yuting Lu1, Yongzhao Shao1. 1. Departments of Population Health and Environmental Medicine, New York University Grossman School of Medicine, New York, NY 10016, USA.
Abstract
Aim: This study aimed to translate a known drug-resistance mechanism of long-term CSF1R inhibition into multicellular biomarkers that can serve as potential therapeutic targets as well as predictive markers for the survival of glioma patients. Methods: Using existing data from a published mouse study of drug resistance in immunotherapy for glioma, we identified multicellular differentially expressed genes (DEGs) between drug-sensitive and drug-resistant mice and translated the DEGs in mouse genome to human homolog. We constructed correlation gene networks for drug resistance in mice and glioma patients and selected candidate genes via concordance analysis of human with mouse gene networks. Markers of drug resistance and an associated predictive signature for patient survival were developed using regularized Cox models with data of glioma patients from The Cancer Genome Atlas (TCGA) database. Predictive performance of the identified predictive signature was evaluated using an independent human dataset from the Chinese Glioma Genome Atlas (CGGA) database. Results: Fourteen genes (CCL22, ADCY2, PDK1, ZFP36, CP, CD2, PLAUR, ACAP1, COL5A1, FAM83D, PBK, FANCA, ANXA7, and TACC3) were identified as genetic biomarkers that were all associated with pathways in glioma progression and drug resistance. Five of the 14 genes (CCL22, ADCY2, PDK1, CD2, and COL5A1) were used to construct a signature that is predictive of patient survival in the proneural subtype GBM patients with an AUC under the time-dependent receiver operating characteristic (ROC) of 2-year survival equal to 0.89. This signature also shows promising predictive accuracy for the survival of LGG patients but not for non-proneural type GBMs. Conclusion: Our translational approach can utilize gene correlation networks from multiple types of cells in the tumor microenvironment of animals. The identified biomarkers of drug resistance have good power to predict patient survival in some major subtypes of gliomas (the proneural subtype of GBM and LGG). The expression levels of the biomarkers of drug resistance may be modified for the development of personalized immunotherapies to prolong survival for a large portion of glioma patients.
Aim: This study aimed to translate a known drug-resistance mechanism of long-term CSF1R inhibition into multicellular biomarkers that can serve as potential therapeutic targets as well as predictive markers for the survival of glioma patients. Methods: Using existing data from a published mouse study of drug resistance in immunotherapy for glioma, we identified multicellular differentially expressed genes (DEGs) between drug-sensitive and drug-resistant mice and translated the DEGs in mouse genome to human homolog. We constructed correlation gene networks for drug resistance in mice and glioma patients and selected candidate genes via concordance analysis of human with mouse gene networks. Markers of drug resistance and an associated predictive signature for patient survival were developed using regularized Cox models with data of glioma patients from The Cancer Genome Atlas (TCGA) database. Predictive performance of the identified predictive signature was evaluated using an independent human dataset from the Chinese Glioma Genome Atlas (CGGA) database. Results: Fourteen genes (CCL22, ADCY2, PDK1, ZFP36, CP, CD2, PLAUR, ACAP1, COL5A1, FAM83D, PBK, FANCA, ANXA7, and TACC3) were identified as genetic biomarkers that were all associated with pathways in glioma progression and drug resistance. Five of the 14 genes (CCL22, ADCY2, PDK1, CD2, and COL5A1) were used to construct a signature that is predictive of patient survival in the proneural subtype GBM patients with an AUC under the time-dependent receiver operating characteristic (ROC) of 2-year survival equal to 0.89. This signature also shows promising predictive accuracy for the survival of LGG patients but not for non-proneural type GBMs. Conclusion: Our translational approach can utilize gene correlation networks from multiple types of cells in the tumor microenvironment of animals. The identified biomarkers of drug resistance have good power to predict patient survival in some major subtypes of gliomas (the proneural subtype of GBM and LGG). The expression levels of the biomarkers of drug resistance may be modified for the development of personalized immunotherapies to prolong survival for a large portion of glioma patients.
Authors: Pedro Berraondo; Miguel F Sanmamed; María C Ochoa; Iñaki Etxeberria; Maria A Aznar; José Luis Pérez-Gracia; María E Rodríguez-Ruiz; Mariano Ponz-Sarvise; Eduardo Castañón; Ignacio Melero Journal: Br J Cancer Date: 2018-11-09 Impact factor: 7.640