Integrating bulk and single-cell sequencing data to construct a Scissor+ dendritic cells prognostic model for predicting prognosis and immune responses in ESCC

Maosheng Cheng, Jianqi Xiong, Qianwen Liu, Caihua Zhang, Kang Li, Xinyuan Wang, Shuang ChenCancer 2024 · DOI

TL;DR

This paper leverages single-cell RNA sequencing (scRNA-seq) and a novel computational method called "Scissor" to identify a specific subtype of dendritic cells (DCs) associated with poor prognosis in esophageal squamous cell carcinoma (ESCC). They developed a prognostic risk model based on these findings, which could lead to more personalized treatment strategies for ESCC patients.

What Problem Does This Paper Address?

Esophageal squamous cell carcinoma (ESCC) is a deadly cancer with significant patient-to-patient variability in how it responds to treatment, largely due to its complex tumor microenvironment (TME). Immune cells, particularly dendritic cells (DCs), play a crucial role in shaping this environment and influencing patient outcomes. However, understanding the specific subtypes of DCs and their precise roles in ESCC has been challenging due to the heterogeneity of the disease. This paper aims to bridge this gap by using advanced single-cell technologies to dissect the DC landscape in ESCC and identify DC subpopulations that are predictive of patient prognosis.

Methods

The researchers employed a multi-pronged approach:

  • scRNA-seq Data Processing and Clustering: They processed a large scRNA-seq dataset to identify and classify different cell types within ESCC tumors. This involved standard dimensionality reduction techniques like PCA and UMAP, followed by clustering to define distinct cell populations.
  • Trajectory Inference: Tools like CytoTRACE and Monocle 2 were used to infer the differentiation pathways of DCs, providing insights into their developmental states.
  • Gene Set Variation Analysis (GSVA): This method was used to analyze the functional enrichment of pathways within different cell populations.
  • Scissor Analysis: This is a key novel method used in the study. Scissor integrates single-cell data with bulk RNA-seq and clinical information (like survival data) to identify cell subpopulations that are specifically associated with a given phenotype (in this case, poor survival). They used Scissor to stratify DCs into "Scissor<sup>hi</sup>" (associated with poor prognosis) and "Scissor<sup>low</sup>" (associated with better prognosis) subtypes.
  • Prognostic Model Construction: Based on the genes identified in the Scissor<sup>hi</sup> DCs, they built a prognostic risk model using LASSO Cox penalized regression. This technique helps select the most relevant genes and build a robust predictive model.
  • Cell Communication Analysis: CellChat was used to investigate the signaling interactions between different cell types within the TME, particularly focusing on differences between Scissor<sup>hi</sup> and Scissor<sup>low</sup> environments.
  • Immunohistochemistry (IHC): This technique was used for clinical validation, allowing researchers to measure the protein expression of key genes identified in the model in actual patient tissue samples.

Datasets

  • ESCC scRNA-seq data (GSE160269): This dataset comprised 192,078 single cells from 60 ESCC patients, providing a high-resolution view of the tumor microenvironment.
  • TCGA clinical data and gene expression: This dataset provided clinical information and bulk RNA-seq data for 79 ESCC patients.
  • ESCC bulk RNA-seq data (GSE53625): This dataset included bulk RNA-seq data from 179 ESCC patients, used for training and validating the Scissor method and the prognostic model.
  • Clinical cohort: A newly collected cohort of 40 ESCC patients was used for the final clinical validation of the prognostic model using immunohistochemistry.

Key Insights and Ideas

  • DC Heterogeneity is Crucial: The study strongly emphasizes that not all DCs are the same. Specific DC subtypes, particularly mature DCs, are significantly associated with worse outcomes in ESCC.
  • Scissor as a Powerful Integrator: The Scissor algorithm proved effective in bridging the gap between single-cell resolution and bulk clinical data, enabling the identification of clinically relevant cell subpopulations.
  • Immune Interactions in Poor Prognosis: Scissor<sup>hi</sup> ESCC patients exhibit distinct cell-cell communication patterns, with increased signaling pathways like PD-L1, TIGIT, PVR, and IL6, suggesting a pro-tumorigenic or immunosuppressive microenvironment.
  • A Six-Gene Prognostic Signature: The researchers successfully distilled the complex DC signature into a concise six-gene panel that can predict patient survival, offering a more practical tool for clinical application.
  • Clinical Relevance Confirmed: The validation using immunohistochemistry on a separate clinical cohort demonstrates that the identified gene signature and the resulting risk model have real-world predictive power.

Key Results

  • Identification of Scissor<sup>hi</sup> DCs: The Scissor analysis identified a subpopulation of DCs (Scissor<sup>hi</sup>) significantly associated with poor overall survival in ESCC. Mature DCs were found to be the predominant cell type within this Scissor<sup>hi</sup> population.
  • Differential Gene Expression: 58 genes were found to be upregulated in Scissor<sup>hi</sup> DCs compared to other DCs.
  • Prognostic Model Performance: A prognostic risk model was developed using six key genes (RPS24, MPP2, TRPM6, SHISA9, CT83, and SPACA4). This model demonstrated significant predictive power for overall survival in both the GEO and TCGA datasets, with AUC values for 1-, 3-, and 5-year survival reaching up to 0.75 in the GEO cohort and 0.97 in the TCGA cohort.
  • Clinical Validation: In a clinical cohort of 40 patients, the risk model's predictions were validated, showing a significant correlation between high-risk scores and poor prognosis.

Limitations and Future Work

  • Limited Sample Size for Clinical Validation: While the clinical validation cohort of 40 patients is valuable, a larger and more diverse cohort would further strengthen the findings.
  • Context-Dependent Role of Mature DCs: The paper notes that mature DCs can have context-dependent roles, correlating with favorable outcomes in some cancers (like ovarian cancer) but poor outcomes in ESCC. This highlights the need for further research into the specific mechanisms driving these differences.
  • Therapeutic Implications: The study suggests that the identified genes and pathways could be targets for personalized therapies, particularly immune checkpoint inhibitors (ICI). Future work could focus on developing and testing targeted therapies based on these findings.
  • Mechanistic Understanding: While the study identifies key genes and interactions, further research into the precise biological mechanisms by which these Scissor<sup>hi</sup> DCs and their associated pathways contribute to ESCC progression is warranted.

Summary updated

Abstract

Esophageal squamous cell carcinoma (ESCC) is characterized by molecular heterogeneity with various immune cell infiltration patterns, which have been associated with therapeutic sensitivity and resistance. In particular, dendritic cells (DCs) are recently discovered to be associated with prognosis and survival in cancer. However, how DCs differ among ESCC patients has not been fully comprehended. Recently, the advance of single-cell RNA sequencing (scRNA-seq) enables us to profile the cell types, states, and lineages in the heterogeneous ESCC tissues. Here, we dissect the ESCC tumor microenvironment at high resolution by integrating 192,078 single cells from 60 patients, including 4379 DCs. We then used Scissor, a method that identifies cell subpopulations from single-cell data that are associated bulk samples with genomic and clinical information, to stratify DCs into Scissor<sup>hi</sup> and Scissor<sup>low</sup> subtypes. We applied the Scissor<sup>hi</sup> gene signature to stratify ESCC scRNAseq patient, and we found that PD-L1, TIGIT, PVR and IL6 ligand-receptor-mediated cell interactions existed mainly in Scissor<sup>hi</sup> patients. Finally, based on the Scissor results, we successfully developed a validated prognostic risk model for ESCC and further validated the reliability of the risk prediction model by recruiting 40 ESCC clinical patients. This information highlights the importance of these genes in assessing patient prognosis and may help in the development of targeted or personalized therapies for ESCC.
BibTeX
@inproceedings{cheng2024integrating,
  author = {Maosheng Cheng and Jianqi Xiong and Qianwen Liu and Caihua Zhang and Kang Li and Xinyuan Wang and Shuang Chen},
  title = {{Integrating bulk and single-cell sequencing data to construct a Scissor+ dendritic cells prognostic model for predicting prognosis and immune responses in ESCC}},
  booktitle = {Cancer},
  year = {2024},
  doi = {10.1007/s00262-024-03683-9},
}
Figure