A T-Cell-Derived 3-Gene Signature Distinguishes SARS-CoV-2 from Common Respiratory Viruses

Li Yang, Xinya Tao, Sheng Ye, Qianchen Tai, Yu-Ang You, Xinting Huang, Mifang Liang, Kai Wang, Haiyan Wen, Chong You, Yan Zhang, Xiao‐Hua ZhouViruses 2024 · DOI

TL;DR

Researchers have identified a novel 3-gene signature (CLSPN, RBBP6, CCDC91) derived from T-cells that can accurately distinguish SARS-CoV-2 infections from other common respiratory viruses, even in asymptomatic cases. This signature holds promise as an early diagnostic tool, capable of detecting SARS-CoV-2 infection before viral RNA is detectable by RT-PCR.

What Problem Does This Paper Address?

The COVID-19 pandemic highlighted the urgent need for rapid and accurate diagnostics, especially given the diverse clinical presentations of SARS-CoV-2 infection, ranging from asymptomatic to severe disease. Distinguishing SARS-CoV-2 from other common respiratory viruses is crucial for effective patient management and public health strategies. While viral RNA detection (like RT-PCR) is standard, it can have a detection window. This paper aims to address this by exploring host-based biomarkers, specifically focusing on the unique immune response patterns elicited by SARS-CoV-2 compared to other prevalent respiratory pathogens like influenza (IFV), respiratory syncytial virus (RSV), and human rhinovirus (HRV). Understanding these host response differences could lead to more robust and potentially earlier diagnostic tools.

Methods

The study employed a multi-pronged approach combining bioinformatics and machine learning. Key methods included:

  • Multi-cohort Analysis: Integrating data from 15 independent cohorts comprising 3730 blood samples.
  • Differential Gene Expression (DEG) Analysis: Identifying genes that are significantly up or down-regulated in infected individuals compared to healthy controls.
  • Gene Ontology (GO) and KEGG Pathway Analysis: Understanding the biological functions and pathways associated with the identified DEGs.
  • Weighted Gene Co-expression Network Analysis (WGCNA): Identifying modules of co-expressed genes and their correlation with disease severity.
  • Machine Learning for Feature Selection: Employing algorithms like Random Forest, XGBoost, GBM, LASSO, and Elastic Net to identify a minimal set of genes that best discriminate SARS-CoV-2.
  • Robust Rank Aggregation (RRA): Aggregating gene importance from multiple machine learning methods to select the most stable candidate genes.
  • Optimal Linear Combination Model: Using a statistical method to derive the best linear combination of selected genes for maximizing diagnostic accuracy (AUC).
  • Single-Cell RNA Sequencing (scRNA-seq): Analyzing scRNA-seq data to pinpoint the specific cell types expressing the identified gene signature.
  • Gene Set Variation Analysis (GSVA): Assessing the enrichment of predefined gene sets (e.g., interferon pathways) across different viral infections.
  • CIBERSORTx: Estimating the abundance of different immune cell types from bulk transcriptomic data.

Datasets

The study utilized a comprehensive collection of publicly available datasets:

  • Discovery Cohorts: HRA000786 and GSE17156 were used for initial discovery and model building.
  • Validation Cohorts: A total of 13 external cohorts were used for validation, including:
  • SARS-CoV-2 specific: GSE157103, GSE198449, GSE157103, GSE201530 (Omicron variant), E-MTAB-10022 (longitudinal follow-up).
  • Other respiratory viruses: GSE68310 (sHCoVs), GSE17156 (IFV, HRV, RSV), GSE61754 (IFV), GSE38900 (IFV, RSV, HRV), GSE67059 (HRV), GSE105450 (RSV).

The study also used scRNA-seq data from the CNP0001102 dataset for cell type analysis.

Key Insights and Ideas

  • Unique SARS-CoV-2 Host Response: SARS-CoV-2 elicits a distinct host immune response compared to other common respiratory viruses. Notably, it shows an enhanced cellular immune response, particularly T-cell activation, but a comparatively limited interferon (IFN) response, especially in asymptomatic cases.
  • T-Cells as Key Players: The identified 3-gene signature is predominantly expressed by T cells, suggesting their critical role in the host's response to SARS-CoV-2. This highlights a potential avenue for understanding disease pathogenesis and developing T-cell-based interventions.
  • The Power of Minimal Signatures: The discovery of a highly accurate 3-gene signature demonstrates that complex biological states can often be captured by a small, well-chosen set of biomarkers, making them more amenable to clinical translation.
  • Early Detection Potential: The signature's ability to predict SARS-CoV-2 infection before viral RNA is detectable by RT-PCR is a significant breakthrough, potentially revolutionizing early diagnosis and containment efforts.
  • Cross-Variant Robustness: The signature's performance with the Omicron variant suggests its potential to remain effective against future SARS-CoV-2 strains.

Key Results

  • A 3-gene signature (CLSPN, RBBP6, CCDC91) was identified that robustly distinguishes SARS-CoV-2 infection from other common respiratory viruses.
  • This signature achieved high diagnostic performance (pooled AUC > 0.9) across multiple independent cohorts, including those with asymptomatic and symptomatic SARS-CoV-2 infections.
  • The signature demonstrated excellent performance in identifying Omicron variant infections (AUC of 0.992).
  • In a longitudinal follow-up study (E-MTAB-10022), the 3-gene signature could identify SARS-CoV-2 infection before detectable viral RNA on RT-PCR (AUC of 0.94).
  • The signature outperformed other published gene expression signatures of similar size in discriminating SARS-CoV-2 infections.
  • The study developed a user-friendly online tool called ViRAL for querying gene expression data across various viral infections.

Limitations and Future Work

The authors acknowledge several limitations:

  • Co-normalization of Discovery Datasets: While necessary for integration, co-normalization might obscure some biological signals. However, the validation in datasets without co-normalization supports the robustness of the signature.
  • Limited Viral Types for Validation: While IFV, RSV, and HRV were included, other common respiratory viruses like human parainfluenza virus (HPIV) and human adenoviruses (HAdV) were not extensively validated.
  • Temporal and Clinical Resolution: The clinical applicability of the signature needs further investigation, especially considering the shortened incubation period of the Omicron variant.
  • Comorbidities: The impact of comorbidities on the signature's performance was not assessed due to data limitations.

Future work suggested includes:

  • Further longitudinal studies to assess the signature's performance across different Omicron variants and over time.
  • Investigating the role of T cells in SARS-CoV-2 pathogenesis and the signature's mechanisms in more detail.
  • Expanding validation to include a wider range of respiratory viruses.
  • Clinical studies to evaluate the signature's utility in real-world diagnostic settings, particularly for early detection and differentiating between viral infections.

Summary updated

Abstract

Research on the host responses to respiratory viruses could help develop effective interventions and therapies against the current and future pandemics from the host perspective. To explore the pathogenesis that distinguishes SARS-CoV-2 infections from other respiratory viruses, we performed a multi-cohort analysis with integrated bioinformatics and machine learning. We collected 3730 blood samples from both asymptomatic and symptomatic individuals infected with SARS-CoV-2, seasonal human coronavirus (sHCoVs), influenza virus (IFV), respiratory syncytial virus (RSV), or human rhinovirus (HRV) across 15 cohorts. First, we identified an enhanced cellular immune response but limited interferon activities in SARS-CoV-2 infection, especially in asymptomatic cases. Second, we identified a SARS-CoV-2-specific 3-gene signature (CLSPN, RBBP6, CCDC91) that was predominantly expressed by T cells, could distinguish SARS-CoV-2 infection, including Omicron, from other common respiratory viruses regardless of symptoms, and was predictive of SARS-CoV-2 infection before detectable viral RNA on RT-PCR testing in a longitude follow-up study. Thereafter, a user-friendly online tool, based on datasets collected here, was developed for querying a gene of interest across multiple viral infections. Our results not only identify a unique host response to the viral pathogenesis in SARS-CoV-2 but also provide insights into developing effective tools against viral pandemics from the host perspective.
BibTeX
@inproceedings{yang2024t,
  author = {Li Yang and Xinya Tao and Sheng Ye and Qianchen Tai and Yu-Ang You and Xinting Huang and Mifang Liang and Kai Wang and Haiyan Wen and Chong You and Yan Zhang and Xiao‐Hua Zhou},
  title = {{A T-Cell-Derived 3-Gene Signature Distinguishes SARS-CoV-2 from Common Respiratory Viruses}},
  booktitle = {Viruses},
  year = {2024},
  doi = {10.3390/v16071029},
}
Figure