Topography of cancer-associated immune cells in human solid tumors

Jakob Nikolas Kather, Meggy Suarez‐Carmona, Pornpimol Charoentong, Cleo‐Aron Weis, Daniela Hirsch, Peter Bankhead, Marcel Horning, Dyke Ferber, Ivan Kel, Esther Herpel, Sarah Schott, Inka Zörnig, Jochen Utikal, Alexander Marx, Timo Gaiser, Hermann Brenner, Jenny Chang‐Claude, Michael Hoffmeister, Dirk Jäger, Niels HalamaeLife 2018 · DOI

TL;DR

This study introduces a systematic approach to classifying immune cell spatial patterns ("topography") within human solid tumors, moving beyond simple cell counts to understand their location relative to the tumor core. They demonstrate that a combined analysis of lymphoid and myeloid cell topographies, rather than individual cell types, can predict patient survival, suggesting a more nuanced approach to cancer immunotherapy biomarkers.

What Problem Does This Paper Address?

The tumor microenvironment is teeming with immune cells that significantly influence cancer progression and response to therapy. While we know these cells are important, understanding their precise spatial organization – their "topography" – within a tumor has been challenging. Previous studies often focused on overall immune cell density or used limited spatial information. This lack of a comprehensive, spatially resolved understanding across different immune cell types and cancer types hinders the development of effective immunotherapies. The authors aim to fill this gap by systematically mapping the distribution of various immune cells within tumors and exploring the clinical relevance of these spatial patterns.

Methods

The core of this study lies in its sophisticated image analysis pipeline. They developed a semiautomatic method to delineate three distinct regions within tumor tissue: the tumor core, the inner invasive margin (500 µm inside the tumor edge), and the outer invasive margin (500 µm outside the tumor edge). Using immunohistochemistry (IHC) to stain for six key immune cell types (CD3, CD8, PD1, Foxp3, CD68, and CD163), they employed an open-source software (QuPath) for automated cell detection and classification. Cell densities within each of the three compartments were then normalized and visualized using "target plots."

Based on these spatial densities, they defined three topographical patterns: "hot" (high density within the tumor core), "cold" (low density throughout), and "immune-excluded" (immune cells primarily at the tumor periphery). Crucially, they moved beyond classifying each cell type independently, exploring bivariate classifications (e.g., CD8 topography combined with CD163 topography) to capture more complex immune phenotypes.

Datasets

The study utilized two main datasets:

  1. Pan-cancer cohort: This dataset comprised 965 histological tissue slides from 177 patients across ten different cancer types. This large, diverse cohort allowed for a broad investigation of immune topographies across various tumor entities.
  2. Colorectal Cancer (CRC) validation cohort (DACHS cohort): An independent cohort of 287 primary CRC patients was used to validate the prognostic significance of the identified immune topographies, specifically focusing on CD8+ lymphocytes and CD163+ macrophages.

Key Insights and Ideas

  • Beyond Density: Location Matters: The study powerfully demonstrates that simply counting immune cells isn't enough. Their spatial arrangement within the tumor microenvironment is critical for understanding tumor immunity and predicting outcomes.
  • A Unified Topographical Framework: They propose a standardized definition of "hot," "cold," and "immune-excluded" topographies that can be applied across multiple immune cell types, providing a common language for describing tumor immune landscapes.
  • Lymphoid and Myeloid Interplay is Key: A major finding is that lymphoid (e.g., CD8+) and myeloid (e.g., CD163+) cell topographies are not always concordant. This discordance is crucial, as a bivariate classification system (combining lymphoid and myeloid patterns) proved to be a much stronger predictor of patient survival than analyzing either cell type alone.
  • Tumor-Specific Immune Signatures: While some general patterns exist, the study highlights that different cancer types exhibit distinct immune topography profiles, suggesting that tumor biology plays a significant role in shaping the immune infiltrate.
  • Prognostic Power of Bivariate Topography: The validation in the CRC cohort is a significant step, showing that a combined CD8/CD163 topography classification can stratify patients for survival, outperforming analyses based on individual cell densities or topographies.

Key Results

  • The pan-cancer analysis revealed that immune topographies vary significantly across different cancer types. For instance, immunogenic tumors like melanoma and lung cancer showed a higher prevalence of "hot" phenotypes, while head and neck tumors often displayed "immune-excluded" patterns.
  • Unsupervised clustering of lymphocyte densities primarily identified "hot" and "cold" tumors, with less clear separation for "immune-excluded" tumors, suggesting that a predefined conceptual framework is useful.
  • In the CRC validation cohort, individual CD8 or CD163 topographies were not prognostic. However, a bivariate classification of CD8 and CD163 topographies identified significant associations with overall survival. Specifically, "CD8-excluded, CD163-excluded" and "CD8-excluded, CD163-hot" phenotypes were associated with poorer survival compared to the "CD8-cold, CD163-cold" reference group.

Limitations and Future Work

  • Subjectivity in Manual Delineation: While the cell detection is automated, the initial manual annotation of the three spatial compartments (tumor core, inner margin, outer margin) could introduce some subjectivity. However, the authors mention reproducibility checks and the use of open-source software for transparency.
  • Limited Cell Types: The study focused on six key immune cell types. Future work could expand this to include a broader range of immune cells and stromal components to further refine the understanding of the tumor microenvironment.
  • Generalizability: While the study is pan-cancer, further validation in a wider array of specific tumor types and across different patient populations would strengthen the findings.
  • Mechanism of Discordance: The study highlights the discordance between lymphoid and myeloid topographies but doesn't deeply explore the underlying biological mechanisms driving these differences. Future research could investigate how tumor cells and other microenvironmental factors orchestrate these distinct spatial patterns.
  • Therapeutic Implications: The authors suggest that this classification could be a novel framework for immunotherapy trials. Future work should focus on correlating these topographies with responses to specific immunotherapies and developing targeted interventions based on these spatial immune phenotypes.

Summary updated

Abstract

Lymphoid and myeloid cells are abundant in the tumor microenvironment, can be quantified by immunohistochemistry and shape the disease course of human solid tumors. Yet, there is no comprehensive understanding of spatial immune infiltration patterns ('topography') across cancer entities and across various immune cell types. In this study, we systematically measure the topography of multiple immune cell types in 965 histological tissue slides from N = 177 patients in a pan-cancer cohort. We provide a definition of inflamed ('hot'), non-inflamed ('cold') and immune excluded patterns and investigate how these patterns differ between immune cell types and between cancer types. In an independent cohort of N = 287 colorectal cancer patients, we show that hot, cold and excluded topographies for effector lymphocytes (CD8) and tumor-associated macrophages (CD163) alone are not prognostic, but that a bivariate classification system can stratify patients. Our study adds evidence to consider immune topographies as biomarkers for patients with solid tumors.
BibTeX
@inproceedings{kather2018topography,
  author = {Jakob Nikolas Kather and Meggy Suarez‐Carmona and Pornpimol Charoentong and Cleo‐Aron Weis and Daniela Hirsch and Peter Bankhead and Marcel Horning and Dyke Ferber and Ivan Kel and Esther Herpel and Sarah Schott and Inka Zörnig and Jochen Utikal and Alexander Marx and Timo Gaiser and Hermann Brenner and Jenny Chang‐Claude and Michael Hoffmeister and Dirk Jäger and Niels Halama},
  title = {{Topography of cancer-associated immune cells in human solid tumors}},
  booktitle = {eLife},
  year = {2018},
  doi = {10.7554/elife.36967},
}
Figure