TISMO: syngeneic mouse tumor database to model tumor immunity and immunotherapy response

Zexian Zeng, Cheryl J. Wong, Lin Yang, Nofal Ouardaoui, Dian Li, Wubing Zhang, Shengqing Gu, Yi Zhang, Yang Liu, Xiaoqing Wang, Jingxin Fu, Liye Zhou, Boning Zhang, Sarah Kim, Kathleen B. Yates, Myles Brown, Gordon J. Freeman, Ravindra Uppaluri, Robert Manguso, X. Shirley LiuNucleic Acids Res 2022 · DOI

TL;DR

Researchers have developed TISMO, a comprehensive database of syngeneic mouse tumor models, to address the challenge of fragmented data in cancer immunology research. This resource provides standardized RNA-seq data, interactive visualization tools, and inferred immune cell infiltration levels, enabling deeper insights into tumor immunity and immunotherapy response.

What Problem Does This Paper Address?

Cancer immunotherapy, particularly immune checkpoint blockade (ICB), has shown remarkable success but only works for a subset of patients. Understanding the complex tumor microenvironment (TME) and why some patients respond while others don't is crucial. While syngeneic mouse models are invaluable for studying these mechanisms in a controlled, immunocompetent setting, the data generated from these models are often scattered across numerous publications, making it difficult to reuse and analyze systematically. This fragmentation hinders the ability to draw broad conclusions and identify common patterns in immunotherapy response. TISMO aims to solve this by consolidating and standardizing this valuable data.

Methods

The TISMO database was built using a multi-step process:

  • Data Collection: Researchers collected RNA-seq data from both public repositories (like GEO) and in-house studies, focusing on syngeneic mouse cancer cell lines and tumor models. They developed a parser to efficiently query and download relevant datasets.
  • Data Standardization: All raw RNA-seq reads were processed using a standardized pipeline called the RNA-seq Immune Analysis Pipeline (RIMA). This pipeline ensures consistency by performing read alignment, quality control, expression quantification, and batch effect removal.
  • Immune Cell Infiltration Inference: To understand the immune landscape within tumors, TISMO integrates inferred immune cell infiltration levels from six state-of-the-art deconvolution algorithms (TIMER, xCell, CIBERSORT, EPIC, quanTIseq, and mMCPcounter). This multi-algorithm approach provides more robust estimations.
  • Metadata Curation: Crucially, detailed phenotypic metadata for each sample, including cancer type, cell line, mouse strain, treatment, and response status, were manually curated from publications.
  • Interactive Web Interface: A user-friendly web interface was developed using the R-Shiny framework, allowing users to explore, visualize, and compare gene expression, pathway enrichment, and immune infiltration levels across different models, treatments, and response groups.

Datasets

TISMO integrates a substantial collection of RNA-seq data:

  • In vitro samples: 605 samples from 49 syngeneic cancer cell lines across 23 cancer types. Of these, 195 were treated with cytokines.
  • In vivo samples: 1518 samples from 68 syngeneic mouse tumor models across 19 cancer types. Notably, 832 of these were from immune checkpoint blockade (ICB) studies.
  • In-house data: An additional 255 in-house RNA-seq samples were included, which had not been previously published.

These datasets represent a significant consolidation of syngeneic mouse model data, differing from existing resources by specifically focusing on expression profiles and providing comprehensive, standardized analysis.

Key Insights and Ideas

  • The Power of Standardization: By uniformly processing vast amounts of data, TISMO enables direct comparisons that were previously challenging due to variations in experimental protocols and analysis methods.
  • Multi-Algorithm Immune Deconvolution: Relying on multiple deconvolution tools for immune cell infiltration provides a more confident and reliable assessment of the immune landscape, mitigating the limitations of any single algorithm.
  • Bridging Pre-clinical and Clinical: Syngeneic models, when analyzed systematically, can serve as powerful platforms to generate and test hypotheses that are directly relevant to human immunotherapy responses.
  • Interactive Exploration: The web interface is a key innovation, allowing researchers to intuitively explore complex relationships between gene expression, immune infiltration, and treatment outcomes without needing extensive bioinformatics expertise.

Key Results

  • TISMO hosts a total of 2123 RNA-seq samples (605 in vitro, 1518 in vivo).
  • The database covers 49 syngeneic cancer cell lines and 68 syngeneic mouse tumor models.
  • A significant portion of the in vivo samples (832) are from ICB studies, making TISMO a valuable resource for understanding ICB response.
  • The interactive web interface allows users to visualize gene expression, pathway enrichment (using ssGSEA), and immune infiltration levels, facilitating hypothesis generation and validation. For example, the paper shows how IFNy expression is upregulated in ICB responders but not non-responders in the CT26 model after ICB treatment.

Limitations and Future Work

  • Data Fragmentation: While TISMO consolidates data, the original studies still represent diverse experimental conditions. The authors acknowledge that comparisons are made within cohorts, and batch effect correction is applied.
  • Model Limitations: Syngeneic models, while powerful, are still simplifications of the human TME. The authors suggest that TISMO will be of "great utility" to the field, implying ongoing value and potential for expansion.
  • Future Work: The paper highlights the ongoing maintenance of TISMO, suggesting that new data and features may be added. The interactive nature of the platform also lends itself to future integration of additional data types or analytical tools.

Summary updated

BibTeX
@article{zeng2022tismo,
  author = {Zexian Zeng and Cheryl J. Wong and Lin Yang and Nofal Ouardaoui and Dian Li and Wubing Zhang and Shengqing Gu and Yi Zhang and Yang Liu and Xiaoqing Wang and Jingxin Fu and Liye Zhou and Boning Zhang and Sarah Kim and Kathleen B. Yates and Myles Brown and Gordon J. Freeman and Ravindra Uppaluri and Robert Manguso and X. Shirley Liu},
  title = {{TISMO: syngeneic mouse tumor database to model tumor immunity and immunotherapy response}},
  journal = {Nucleic Acids Res.},
  year = {2022},
  doi = {10.1093/nar/gkab804},
}
Figure