P72.02 Cellular Landscape of Tumor Immune Microenvironment and Genetic Signatures Identify Prognostic of LUAD

Abstract

Tumour microenvironment (TME) has been recognized to support the initiation and progression of lung adenocarcinoma (LUAD). The innate and adaptive immune cells in the lung TME harbour both tumour-promoting and tumour-suppressing activities, which may also predict clinical outcome. Therefore we carried out a systematic analysis of cellular interactions in tumor immune microenvironment. And identify cell-intrinsic and cell-extrinsic pathways cell types and activation states that may serve as biomarkers of overall survival (OS). Public gene-expression data and relevant clinical annotation were obtained from Gene-Expression Omnibus (GEO) and the Cancer Genome Atlas (TCGA) database. Three TME infiltration patterns were comprehensively analyzed in 442 LUAD patients using CIBERSORT algorithm and the LM22 gene signature. Based on the TME patterns, we build a model to calculate TMEscore based on gene set variation analysis via ssGSEA algorithm. Functional enrichment analysis were performed by GO and KEGG. Four datasets with available outcome data and clinical information in GEO and TCGA-LUAD were enrolled in our study. GSE72094 was used as the training cohort, while GSE11969, GSE26939, GSE31210 and TCGA-LUAD was used as validation cohorts. TME cell network established based on GSE72094 depicted a comprehensive landscape of tumor-immune cell interactions, cell lineages, and their correlation with OS (Fig. 1A, 1B). Three subgroups with distinct TME signature gene sets were obtained/identified based on unsupervised hierarchical clustering in 442 LUAD cases. OS in TME gene subgroup B was significantly longer than which in TME gene subgroup A and subgroup C. TME gene group B was associated immune activation (Fig. 1C). TMEscore was further constructed using principal component analysis algorithms. Lower TMEscore is significantly associated with better prognosis. Functional annotation analysis showed TMEscore had a positive correlation with cell cycle, DNA replication, homologous recombination, mismatch repair, nucleotide excision repair and DNA damage repair (Fig. 1D). The enriched pathways in subtype with lowest/low TMEscore involved bile_acid_metabolism, fatty_acid_metabolism and myogenesis. While high TMEscore subtype was characterized by significant enrichment of interferon_alpha_response, myc_targets and unfolded_protein_response pathway (Fig. 1E). TMEscore model was then validated on 525 patients from GEO datasets and 585 patients from TCGA-LUAD project and proved to be a valuable method for prognostic stratification of LUAD except for TNM stage (Fig. 1F). Variability in the composition of the tumor immune microenvironment contributes to heterogeneity in OS. Deeper validation is in need to define the positive association between lower TMEscore and longer OS.