An integrated bioinformatic analysis of bulk and single-cell sequencing clarifies immune microenvironment and metabolic profiles of lung adenocarcinoma to predict immunotherapy efficacy

Abstract

Targeting the tumor microenvironment is increasingly recognized as an effective treatment of advanced lung adenocarcinoma (LUAD). However, few studies have addressed the efficacy of immunotherapy for LUAD. Here, a novel method for predicting immunotherapy efficacy has been proposed, which combines single-cell and bulk sequencing to characterize the immune microenvironment and metabolic profile of LUAD. TCGA bulk dataset was used to cluster two immune subtypes: C1 with "cold" tumor characteristics and C2 with "hot" tumor characteristics, with different prognosis. The Scissor algorithm, which is based on these two immune subtypes, identified GSE131907 single cell dataset into two groups of epithelial cells, labeled as Scissor_C1 and Scissor_C2. The enrichment revealed that Scissor_C1 was characterized by hypoxia, and a hypoxic microenvironment is a potential inducing factor for tumor invasion, metastasis, and immune therapy non-response. Furthermore, single cell analysis was performed to investigate the molecular mechanism of hypoxic microenvironment-induced invasion, metastasis, and immune therapy non-response in LUAD. Notably, Scissor_C1 cells significantly interacted with T cells and cancer-associated fibroblasts (CAF), and exhibited epithelial-mesenchymal transition and immunosuppressive features. CellChat analysis revealed that a hypoxic microenvironment in Scissor_C1elevated TGFβ signaling and induced ANGPTL4 and SEMA3C secretion. Interaction with endothelial cells with ANGPTL4, which increases vascular permeability and achieves distant metastasis across the vascular endothelium. Additionally, interaction of tumor-associated macrophages (TAM) and Scissor_C1 via the EREG/EFGR pathway induces tyrosine kinase inhibitor drug-resistance in patients with LAUD. Thereafter, a subgroup of CAF cells that exhibited same features as those of Scissor_C1 that exert immunosuppressive functions in the tumor microenvironment were identified. Moreover, the key genes (EPHB2 and COL1A1) in the Scissor_C1 gene network were explored and their expressions were verified using immunohistochemistry. Finally, the metabolism dysfunction in cells crosstalk was determined, which is characterized by glutamine secretion by TAM and uptake by Scissor_C1 via SLC38A2 transporter, which may induce glutamine addiction in LUAD cells. Overall, single-cell sequencing clarifies how the tumor microenvironment affects immunotherapy efficacy via molecular mechanisms and biological processes, whereas bulk sequencing explains immunotherapy efficacy based on clinical information.

Keywords: cancer-associated fibroblasts; epithelia-mesenchymal transition; immunotherapy; lung adenocarcinoma; metabolism; single-cell multi-omics.

FIGURE 1

Overview of the work in this study.

FIGURE 2

Identification of two immune subtypes based on TCGA-LUAD (A) Identification of two immune subtypes, C1 and C2, based on 510 samples with NMF optimal k = 2. (B) GSEA analysis showed that C2 had higher immune activity compared to C1. (C) Relative abundance of immune cells in both immune subtypes. (D) Heatmap of the 29 immunogenomes of the two immune subtypes, with higher cytotoxicity, immune checkpoint and pro-inflammatory process activity in C2 compared to C1. (E) Tumor microenvironment profiles of the two immune subtypes.

FIGURE 3

(A) Prognostic value of both immune subtypes, C1 subtype had poor prognosis compared to C2, and C1 immunotherapy response score was lower. (B) Immunotherapy efficacy (EaSIeR score and PD-L1 expression) (C) Resistance of two immune subtypes to EGFR-TKI drugs.

FIGURE 4

Characterization of two groups of Scissor epithelial cells (A,B) The Scissor algorithm identifies the two epithelial cells most associated with the two immune subtypes, Scissor_C1 and Scissor_C2 cells. (C) Scissor_C1 significantly up- and downregulated genes. (D) KEGG and GSEA enrichment analysis of Scissor_C1 revealing hypoxia and EMT related biological processes (Focal adhesion, ECM-receptor interaction and HIF-1 signaling pathway. (E) Revealing Scissor_C2 enrichment in oxidative phosphorylation and immune activity related pathways (antigen presentation, Th17 cell differention).

FIGURE 5

(A,B) Prognostic analysis of two groups of Scissor cells deconvoluted to bulk sequencing (C,D) Scissor_C1 cells exhibit different abundance in stage and N-stage deconvolution. (E,F) Scissor_C1 was highly expressed in the immunotherapy non-responsive group.

FIGURE 6

Cellular communication networks of two groups of Scissor cells (A) Communication between Scissor cells and immune, stromal cell growth factors. (B) Communication between Scissor cells and MHC-like molecules of immune, stromal cells. (C) Communication between Scissor cells and immune, stromal cell colony-stimulating factors (D) TGFβ analysis and cell growth factor communication.

FIGURE 7

(A–F) ssGSEA infiltration scores of T-cells in two groups of immune subtypes. (D,G,H) Correlation analysis of Scissor cells with T-cells in both groups.

FIGURE 8

(A) Scissor and stromal cell signaling-related pathways. (B) Scissor and stromal cell release signaling-related pathways. (C) Scissor_C1 cells interact with CAF, endothelial cells through the release of ANGPTL4. (D) Scissor_C1 releases SEMA3C to interacts with endothelial cells.

FIGURE 9

(A) Myeloid cell descending clustering and cell annotation, characteristic marker gene expression. (B) Signaling reception-related pathways in Scissor and TAM cells. (C) Scissor cell and TAM cell release signaling-related pathways. (D) Interaction of Scissor_ C1-released ANGPTL4 with TAM cells. (E) Scissor_C1 cells release SEMA3C to interact with TAM cells. (F) TAM interacts with Scissor_C1 via the EGF signaling pathway. (G) Scissor_C1 interacts with TAM cells through the complement pathway.

FIGURE 10

Results of Scissor identification in CAF cells (A) Visualization of UAMP in CAF cells, caf_C1_scissor for Scissor cells related to C1 subtype, and caf_C2_scissor for Scissor cells related to C2 subtype. (B) Volcano plot of differential expression of caf_C1_scissor and caf_C2_scissor. (C) Map of GSVA pathway in CAF-related Scissor cells. (D,G) caf_C1_scissor interacts significantly with TAM and T-cells via CXCL12_CXCR4. (E) caf_C1_scissor1 interacts significantly with C1QA + TAM via IGF signaling pathway. (F) caf_C1_scissor interacts significantly with TAM cells through the POSTN signaling pathway. (H) caf_C1_scissor interacts significantly with Treg cells via IL16 signaling pathway.

FIGURE 11

(A) Transcription factors significantly differentially activated by Scissor_C1 and Scissor_C2. (B) Scissor_C1-specific gene regulatory network. (C) GSEA analysis of high and low expression of COL1A1. (D) Scissor_C1 is specifically activated by ligand pairs with EPHB2-EFNB1 in endothelial cells. (E) Prognostic comparison of COL1A1 and EPHB2. (F–H) EPHB2 is significantly and positively correlated with immunosuppressive markers PD-1,LAG3,TIGIT.

FIGURE 12

Immunohistochemistry (A)COL1A1 expressed in tumor tissues (B) COL1A1 expressed in normal tissues (C)EPHB2 expressed in tumor tissues (D) EPHB2 expressed in normal tissues.

FIGURE 13

(A) TAM cells secreting glutamine metabolites bind to the SLC38A2 transporter protein on Scissor_C1. (B) TAM cells secreting glutamine metabolites bind to SLC3A2 and SLC38A2 transporter proteins on Scissor_C2 cells. (C) Scissor_C2 is more active in metabolite-based communication with TAM compared to Scissor_C1. (D) Both C1 and C2 subtypes show significant differences in metabolic pathway activation, with C2 subtype showing significant metabolic pathway activation.

FIGURE 14

Mechanism of invasive metastasis and immunosuppressive effect of Scissor_C1.