A prognostic and therapeutic hallmark developed by the integrated profile of basement membrane and immune infiltrative landscape in lung adenocarcinoma

Abstract

Basement membranes (BMs) are specialised extracellular matrices that maintain cellular integrity and resist the breaching of carcinoma cells for metastases while regulating tumour immunity. The tumour immune microenvironment (TME) is essential for tumour growth and the response to and benefits from immunotherapy. In this study, the BM score and TME score were constructed based on the expression signatures of BM-related genes and the presence of immune cells in lung adenocarcinoma (LUAD), respectively. Subsequently, the BM-TME classifier was developed with the combination of BM score and TME score for accurate prognostic prediction. Further, Kaplan-Meier survival estimation, univariate Cox regression analysis and receiver operating characteristic curves were used to cross-validate and elucidate the prognostic prediction value of the BM-TME classifier in several cohorts. Findings from functional annotation analysis suggested that the potential molecular regulatory mechanisms of the BM-TME classifier were closely related to the cell cycle, mitosis and DNA replication pathways. Additionally, the guiding value of the treatment strategy of the BM-TME classifier for LUAD was determined. Future clinical disease management may benefit from the findings of our research.

Keywords: basement membrane; chemotherapy drug screening; immune infiltration; immunotherapy response; lung adenocarcinoma; prognosis.

Figure 1

The workflow for screening basement membrane-related genes and immune cell types for establishing the BM score and TME score. (A) Differential expression analysis, univariate Cox regression analysis and lasso regression analysis were performed for 160 BM-related genes in TCGA-LUAD cohort in succession. The upper line marked the number of BM-related genes before and after each step analysis. Briefly, 20 BM-related genes were selected for the development of BM score. (B) The CIBERSORT algorithm was applied to generate the abundance of 22 types of immune cells for TCGA-LUAD samples. After abandoning 3 types of immune cells with few abundances, Kaplan-Meier survival curve estimation was executed for the remaining immune cell types. Kaplan-Meier overall survival curves were shown on the right side only for immune cell types with p-values less than 0.05 (purple indicated favorable prognostic factor while red represented unfavorable prognostic factor). Ultimately, mast cells resting, monocytes and plasma cells were used for the establishment of TME score.

Figure 2

Development and performance of BM and TME scores in LUAD, respectively. (A, B) The correlation of BM score and TME score with the expression levels of prognostically BM genes and the abundance of prognostically immune cells. (C, D) Kaplan–Meier survival curves of high and low BM, TME scores subgroups. (E, F) Top three KEGG enrichment pathways based on GSEA analysis of differentially expressed genes (DEGs) between high and low BM, TME scores groups.

Figure 3

The relationship between BM score and different types of cells and tumours. (A) t-SNE scatters plot of eight primary and five metastatic LUAD samples with cell annotation. (B) The distribution of BM scores at the single cell level. Blue and red circles represent epithelial and stromal cells, respectively. The rest of the scatter plot is filled with immune cells. (C) The violin plot demonstrating the difference in BM score among immune, stromal and epithelial cells. And the red dots indicate the average value of BM scores for each group. (D) The relationship between BM score and the abundance of T cells CD4 memory resting in TCGA-LUAD cohort. (E) The violin plot showing the difference in BM score between primary and metastatic LUAD tumour cells. Further, the red dots indicate the average value of BM scores for each group. (F) The distribution of BM score between tumours with non-metastatic (M0) and metastatic (M1). **p<=0.01 and ***p<=0.001, respectively.

Figure 4

Prognostic value and enrichment analysis relevant to BM-TME classifier. (A–C) Kaplan–Meier overall survival curves of the training set (TCGA-LUAD cohort) and validation sets (GSE30219 and GSE81089 cohort) based on BM-TME classifier. (D) Heat map depicting the correlation between various gene modules of WGCNA analysis and BM-TME subgroups. (E) Top ten biological process (BP), cellular component (CC) and molecular function (MF) enrichment pathways based on GO analysis of blue module genes. Bar plots correspond to the lower axis; dot plots correspond to the upper axis.

Figure 5

Relationships between BM-TME classifier and clinical features in LUAD. (A) ROC curves for the 3-, 5- and 7-year overall survival based on the BM-TME classifier in TCGA-LUAD cohort. (B) Univariate Cox analysis of clinical characteristics and BM-TME classifier in five LUAD cohorts. (C–J) Kaplan–Meier overall survival curves of BM-TME classifier in diverse LUAD clinical subtypes in TCGA-LUAD cohort.

Figure 6

Correlations of immune checkpoints and somatic mutation with BM-TME classifier in TCGA-LUAD cohort. (A) The differential expression levels of immune checkpoint genes among BM-TME classifier subgroups. (B) Waterfall plots depicting the mutation landscape of the top 15 genes with high mutation frequency. P-values on the right side displaying the significance of differences in 15 gene mutation frequencies between two BM-TME groups. (C) Kaplan–Meier curves of patients with LUAD divided by the CSMD3 mutation status and BM-TME classifier. (D) The distribution of TMB among different BM-TME classifier subgroups. *p<=0.05, **p<=0.01 and ***p<=0.001, respectively. ns, no significance.

Figure 7

Immunotherapy response and chemotherapy drug screening prediction. (A, B) Differential distribution of BM score and TME score in immunotherapy response and non-response groups. (C) Comparison of immunotherapy responses among different BM-TME classifier groups in TCGA-LUAD cohort. (D, E) Proteomaps of the functional analysis results in patients of “BM_low+TME_high”, “BM_high+TME_low”, immunotherapy responder and non-responder groups. Each KEGG pathway is represented by a polygon, and the size of polygons corresponds to the protein ratio. (F–H) Comparison of drug sensitivity to cisplatin, paclitaxel and vinorelbine among different BM-TME subgroups. *p<=0.05, **p<=0.01 and ***p<=0.001, respectively. ns, no significance.