Machine learning identifies characteristics molecules of cancer associated fibroblasts significantly correlated with the prognosis, immunotherapy response and immune microenvironment in lung adenocarcinoma

Qian Wang; Xunlang Zhang; Kangming Du; Xinhui Wu; Yexin Zhou; Diang Chen; Lin Zeng

doi:10.3389/fonc.2022.1059253

Machine learning identifies characteristics molecules of cancer associated fibroblasts significantly correlated with the prognosis, immunotherapy response and immune microenvironment in lung adenocarcinoma

Front Oncol. 2022 Nov 9:12:1059253. doi: 10.3389/fonc.2022.1059253. eCollection 2022.

Authors

Qian Wang¹, Xunlang Zhang², Kangming Du³, Xinhui Wu², Yexin Zhou⁴, Diang Chen⁵, Lin Zeng⁶

Affiliations

¹ Hospital of Chengdu University of Traditional Chinese Medicine, Chengdu, Sichuan, China.
² Department of Geriatric, Hospital of Chengdu University of Traditional Chinese Medicine, Chengdu, Sichuan, China.
³ Department of Vascular Surgery, Hospital of Chengdu University of Traditional Chinese Medicine, Chengdu, Sichuan, China.
⁴ Guangxi University of Chinese Medicine, Nanning, China.
⁵ Department of Urology Surgery, Hospital of Chengdu University of Traditional Chinese Medicine, Chengdu, Sichuan, China.
⁶ Department of Neurosurgery, Hospital of Chengdu University of Traditional Chinese Medicine, Chengdu, Sichuan, China.

Abstract

Background: Lung adenocarcinoma (LUAD) is a highly lethal disease with a dramatic pro-fibrocytic response. Cancer-associated fibroblasts (CAFs) have been reported to play a key role in lung adenocarcinoma.

Methods: Marker genes of CAFs were obtained from the Cell Marker website. Single sample gene set enrichment analysis (ssGSEA) was used for CAFs quantification. R and GraphPad Prism software were utilized for all analysis. Quantitative real-time PCR (qRT-PCR) was utilized to detect the RNA level of specific molecules.

Results: Based on the ssGSEA algorithm and obtained CAFs markers, the LUAD patients with low- and high-CAFs infiltration were successfully identified, which had different response patterns to immunotherapy. Through the machine learning algorithm - LASSO logistic regression, we identified 44 characteristic molecules of CAFs. Furthermore, a prognosis signature consisting of seven characteristic genes was established, which showed great prognosis prediction ability. Additionally, we found that patients in the low-risk group might have better outcomes when receiving immunotherapy of PD-1, but not CTLA4. Also, the biological enrichment analysis revealed that immune response-related pathways were significantly associated with CAFs infiltration. Meanwhile, we investigated the underlying biological and microenvironment difference in patients with high- and low-risk groups. Finally, we identified that AMPD1 might be a novel target for LUAD immunotherapy. Patients with a high level of AMPD1 were correlated with worse responses to immunotherapy. Moreover, immunohistochemistry showed that the protein level of AMPD1 was higher in lung cancer. Results of qRT-PCR demonstrated that AMPD1 was upregulated in A549 cells compared with BEAS-2B. Meanwhile, we found that the knockdown of AMPD4 can significantly reduce the expression of CTLA4 and PDCD1, but not CD274 and PDCD1LG2.

Conclusion: We comprehensively explored the role of CAFs and its characteristics molecules in LUAD immunotherapy and developed an effective signature to indicate patients prognosis and immunotherapy response. Moreover, AMPD1 was identified as a novel target for lung cancer immunotherapy.

Keywords: TIDE; cancer-associated fibroblasts; immune microenvironment; immunotherapy; lung adenocarcinoma.