Spatial Deconvolution of HER2-positive Breast Tumors Reveals Novel Intercellular Relationships

Alma Andersson, Ludvig Larsson, Linnea Stenbeck, Fredrik Salmén, Anna Ehinger, Sunny Z. Wu, Ghamdan Al‐Eryani, Daniel Roden, Alexander Swarbrick, Åke Borg, Jonas Frisén, Camilla Engblom, Joakim LundebergbioRxiv 2020 · DOI

TL;DR

This study uses Spatial Transcriptomics (ST) to map gene expression and cell types within HER2-positive breast tumors, revealing intricate spatial relationships between different cell populations. By integrating ST with single-cell RNA sequencing (scRNA-seq), the researchers uncover novel intercellular interactions and develop a predictive model for tertiary lymphoid structures (TLSs), offering new insights into tumor microenvironments and potential therapeutic targets.

What Problem Does This Paper Address?

Traditional transcriptomic studies, while powerful for understanding gene expression, often sacrifice crucial spatial information. This loss of context makes it difficult to decipher how cells interact within a tissue, a critical factor in understanding cancer development and treatment response. Specifically, for HER2-positive breast tumors, a significant subtype with aggressive growth, understanding the complex interplay of immune cells and tumor cells within their spatial microenvironment is essential for developing more effective therapies. Existing methods like immunohistochemistry (IHC) and in situ hybridization (ISH) offer spatial resolution but require prior knowledge of specific targets, limiting exploratory analysis. Single-cell RNA sequencing (scRNA-seq) provides high-throughput transcriptomic data but loses spatial origin. Spatial Transcriptomics (ST) bridges this gap by providing spatially resolved, transcriptome-wide expression data, but it lacks single-cell resolution. This paper aims to overcome these limitations by integrating ST with scRNA-seq to map cell types and their spatial organization within HER2-positive breast tumors, thereby uncovering novel intercellular relationships and identifying key structures like tertiary lymphoid structures (TLSs).

Methods

The study employs a multi-faceted approach combining advanced techniques:

  • Spatial Transcriptomics (ST): This core technology allows for the measurement of gene expression in spatially defined locations within tissue sections. The researchers used ST on 36 sections from eight HER2-positive breast tumors.
  • Expression-based Clustering: Unsupervised clustering algorithms were applied to the ST data to group spots with similar gene expression profiles. This data-driven approach helps identify distinct cellular neighborhoods within the tumor.
  • Single-Cell RNA Sequencing (scRNA-seq) Integration: scRNA-seq data from HER2-positive tumors was used to deconvolve the spatial data. This means mapping the identified cell types from scRNA-seq onto the spatial spots from ST, providing a higher resolution understanding of cell composition within each spatial location. The stereoscope package was used for this integration, employing a probabilistic model.
  • Functional Enrichment Analysis: Marker genes identified for each cluster were analyzed using Gene Ontology (GO) databases to understand the biological pathways and functions associated with these distinct spatial regions.
  • Cell Type Co-localization Analysis: Pearson correlation was used to quantify the spatial co-occurrence of different cell types, revealing which cell populations tend to be found together within the tumor microenvironment.
  • Tertiary Lymphoid Structure (TLS) Inference and Prediction: A novel method was developed to identify potential TLSs by analyzing the co-localization of B and T-cells. A TLS-score was computed, and a linear model was trained to predict TLS presence based on gene expression, creating a "TLS-signature." This model was then validated on external datasets.

Datasets

The study utilized:

  • 36 ST sections from eight HER2-positive breast tumor individuals.
  • Single-cell RNA sequencing (scRNA-seq) data from five HER2-positive tumors.
  • External datasets for model validation, including breast cancer data from a different platform (Visium), as well as tissues from rheumatoid arthritis, melanoma, and developmental heart.

Key Insights and Ideas

  • Spatial Heterogeneity is Key: The study highlights that HER2-positive breast tumors are not uniform but exhibit significant spatial heterogeneity in gene expression and cell composition, both within and between patients. This underscores the importance of spatial context in understanding tumor biology.
  • Immune Microenvironment Matters: The research reveals distinct spatial patterns of immune cell infiltration and interaction within the tumor. This includes co-localization of B and T-cells, and macrophage-T-cell subsets, suggesting active immune surveillance or modulation within the tumor.
  • Tertiary Lymphoid Structures (TLSs) as Potential Therapeutic Targets: The identification and predictive modeling of TLSs is a significant finding. TLSs are known to play a role in anti-tumor immunity, and their presence can be linked to treatment outcomes. This work provides a method to identify and characterize these structures spatially.
  • Integration is Powerful: The successful integration of ST and scRNA-seq data demonstrates the power of combining different omics technologies to gain a more comprehensive understanding of complex biological systems like the tumor microenvironment.

Key Results

  • Data-driven Annotation: Expression-based clustering successfully identified distinct spatial regions within tumors, which largely corresponded to pathologist-defined morphological regions, but also revealed finer distinctions.
  • Shared Gene Signatures: The study identified shared gene signatures for immune and tumor processes across HER2-positive patients, suggesting common underlying mechanisms.
  • Cell Type Co-localization Patterns: Novel patterns of cell type co-localization were discovered, including segregated epithelial cells, interactions between B and T-cells, and co-localization of macrophages and T-cell subsets.
  • TLS Prediction Model: A predictive model for TLS presence was developed, demonstrating generalizability across different tissues and platforms, suggesting its potential as a biomarker or therapeutic target.
  • Type I Interferon Response: The study found evidence of spatially restricted type I interferon responses in some HER2-positive tumors, potentially involving macrophage-T-cell interactions.

Limitations and Future Work

  • Single-Cell Resolution in ST: While ST provides spatial information, it does not achieve true single-cell resolution. The integration with scRNA-seq helps to deconvolve this, but limitations remain in distinguishing very rare cell types or subtle differences within closely related cell populations.
  • External Data Limitations: The study acknowledges that single-cell data used for deconvolution might not perfectly match the spatial data, potentially leading to limitations in inferring proportions of certain cell types (e.g., neutrophils and mast cells were not included due to their absence in the single-cell dataset).
  • Generalizability of TLS Model: While the TLS prediction model showed promise on external data, further studies are needed to confirm its broad applicability and clinical relevance.
  • Future Directions: The authors suggest future work could link cell co-localization patterns to patient outcomes, assess drug responses in a spatially restricted manner, and further investigate the functional interactions between cell types, particularly the synergistic relationship between macrophages and T-cells.

Summary updated

Abstract

ABSTRACT In the past decades, transcriptomic studies have revolutionized cancer treatment and diagnosis. However, tumor sequencing strategies typically result in loss of spatial information, critical to understand cell interactions and their functional relevance. To address this, we investigate spatial gene expression in HER2-positive breast tumors using Spatial Transcriptomics technology. We show that expression-based clustering enables data-driven tumor annotation and assessment of intra-and interpatient heterogeneity; from which we discover shared gene signatures for immune and tumor processes. We integrate and spatially map tumor-associated types from single cell data to find: segregated epithelial cells, interactions between B and T-cells and myeloid cells, co-localization of macrophage and T-cell subsets. A model is constructed to infer presence of tertiary lymphoid structures, applicable across tissue types and technical platforms. Taken together, we combine different data modalities to define novel interactions between tumor-infiltrating cells in breast cancer and provide tools generalizing across tissues and diseases.
BibTeX
@misc{andersson2020spatial,
  author = {Alma Andersson and Ludvig Larsson and Linnea Stenbeck and Fredrik Salmén and Anna Ehinger and Sunny Z. Wu and Ghamdan Al‐Eryani and Daniel Roden and Alexander Swarbrick and Åke Borg and Jonas Frisén and Camilla Engblom and Joakim Lundeberg},
  title = {{Spatial Deconvolution of HER2-positive Breast Tumors Reveals Novel Intercellular Relationships}},
  year = {2020},
  doi = {10.1101/2020.07.14.200600},
  howpublished = {bioRxiv},
}
Figure