Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data.
Racle J, de Jonge K, Baumgaertner P, Speiser DE, Gfeller D. · 2017
Read Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data. by Racle J, de Jonge K, Baumgaertner P, Speiser DE, Gfeller D. (2017) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5718706.txt.
Introduction
Racle and co-authors present a method to count cell types inside tumours. Their research article appeared in eLife. Bulk tumour gene data mix many cell types. Immune cells in that mix shape outcomes. Their algorithm estimates each cell fraction. It covers cancer and immune cells at once.
The method uses fresh expression profiles. Each major non-cancer cell type has one. It corrects for cell-specific mRNA content. Unknown cell types can be included too. The team validated the method three ways. Flow cytometry was one check. Tissue staining and single-cell data were others. Human melanoma samples were tested. Colorectal tumours were tested as well. The tool is public online.
Catalog record
| CREATOR | Racle J, de Jonge K, Baumgaertner P, Speiser DE, Gfeller D. |
|---|---|
| YEAR | 2017 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 74 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data By Julien Racle, Kaat de Jonge, Petra Baumgaertner, Daniel E Speiser, David Gfeller, Alfonso Valencia Published in: eLife DOI: 10.7554/eLife.26476 ABSTRACT Immune cells infiltrating tumors can have important impact on tumor progression and response to therapy. We present an efficient algorithm to simultaneously estimate the fraction of cancer and immune cell types from bulk tumor gene expression data. Our method integrates novel gene expression profiles from each major non-malignant cell type found in tumors, renormalization based on cell-type-specific mRNA content, and the ability to consider uncharacterized and possibly highly variable cell types. Feasibility is demonstrated by validation with flow cytometry, immunohistochemistry and single-cell RNA-Seq analyses of human melanoma and colorectal tumor specimens. Altogether, our work not only improves accuracy but also broadens the scope of absolute cell fraction predictions from tumor gene expression data, and provides a unique novel experimental benchmark for immunogenomics analyses in cancer research (http://epic.gfellerlab.org). INTRODUCTION Tumors form complex microenvironments composed of various cell types such as cancer, immune, stromal and endothelial cells (Hanahan and Weinberg, 2011; Joyce and Fearon, 2015). Immune cells infiltrating the tumor microenvironment play a major role in shaping tumor progression, response to (immuno-)therapy and patient survival (Fridman et al., 2012). Today, gene expression analysis is widely used to characterize tumors at the molecular level. As a consequence, tumor gene expression profiles from tens of thousands of patients are available across all major tumor types in databases such as Gene Expression Omnibus (GEO [Edgar et al., 2002]) or The Cancer Genome Atlas (TCGA [Hoadley et al., 2014]). Unfortunately, flow cytometry or immunohistochemistry (IHC) measurements to quantify the number of both malignant and tumor-infiltrating immune cells are rarely performed for samples analyzed at the gene expression level. Therefore, to correctly interpret these data in particular from an immuno-oncology point of view (Angelova et al., 2015; Gentles et al., 2015; Hackl et al., 2016; Li et al., 2016; Linsley et al., 2015; Rooney et al., 2015; Şenbabaoğlu et al., 2016; Zheng et al., 2017), reliable and carefully validated bioinformatics tools are required to infer the fraction of cancer and immune cell types from bulk tumor gene expression data. To this end, diverse bioinformatics methods have been developed. Some aim at estimating tumor purity based on copy number variation (Carter et al., 2012; Li and Li, 2014), or expression data (Ahn et al., 2013; Clarke et al., 2010; Quon et al., 2013; Yoshihara et al., 2013), but do not provide information about the different immune cell types. Others focus on predicting the relative proportions of cell types by fitting reference gene expression profiles from sorted cells (Gong and Szustakowski, 2013; Li et al., 2016; Newman et al., 2015; Qiao et al., 2012) or with the help of gene signatures (Becht et al., 2016; Zhong et al., 2013). These approaches have been recently applied to cancer genomics data to investigate the influence of immune infiltrates on survival or response to therapy (Charoentong et al., 2017; Gentles et al., 2015; Şenbabaoğlu et al., 2016) or predict potential targets for cancer immunotherapy (Angelova et al., 2015; Li et al., 2016). However, none of these methods provides quantitative information about both cancer and non-malignant cell type proportions directly from tumor gene expression profiles. In addition, reference gene expression profiles used in previous studies have been mainly obtained from circulating immune cells sorted from peripheral blood and were generally based on microarrays technology. Finally, several of these approaches have not been experimentally validated in solid tumors from human patients. Here, we developed a robust approach to simultaneously Estimate the Proportion of Immune and Cancer cells (EPIC) from bulk tumor gene expression data. EPIC is based on a unique collection of RNA-Seq reference gene expression profiles from either circulating immune cells or tumor- infiltrating non-malignant cell types (i.e., immune, stromal and endothelial cells). To account for the high variability of cancer cells across patients and tissue of origin, we implemented in our algorithm the ability to consider uncharacterized, possibly highly variable, cell types. To validate our predictions in human solid tumors, we first analyzed melanoma samples with both flow cytometry and RNA-Seq. We then collected publicly available IHC and single-cell RNA-Seq data of colorectal and melanoma tumors. All three validation datasets showed that very accurate predictions of both cancer and non-malignant cell type proportions could be obtained even in the absence of a priori information about cancer cells. RESULTS Reference gene expression profiles from circulating and tumor-infiltrating cells EPIC incorporates reference gene expression profiles from each major immune and other non-malignant cell type to model bulk RNA-Seq data as a superposition of these reference profiles (Figure 1A,B). To tailor our predictions to recent gene expression studies, we first collected and curated RNA-Seq profiles of various human innate and adaptive circulating immune cell types (Hoek et al., 2015; Linsley et al., 2014; Pabst et al., 2016) (CD4 T cells, CD8 T cells, B cells, NK cells, Monocytes and Neutrophils) from a diverse set of patients analyzed in different centers (see Materials and methods). Principal component analysis (PCA) of these data (Figure 1C) showed that samples clustered first according to cell type and not according to experiment of origin, patient age, disease status or other factors, suggesting that they could be used as bona fide reference expression profiles across different patients. Reference gene expression profiles for each major immune cell type were built from these RNA-Seq samples based on the median normalized counts per gene and cell type. The variability in expression for each gene was also considered when predicting the various cell proportions based on these reference profiles (see Materials and methods and Supplementary file 1). [Figure: Estimating the proportion of immune and cancer cells.(A) Schematic description of our method. (B) Matrix formulation of our algorithm, including the uncharacterized cell types (red box) with no or very low expression of signature genes (green box). (C) Low dimensionality representation (PCA based on the 1000 most variable genes) of the samples used to build the reference gene expression profiles from circulating immune cells (study 1 [Hoek et al., 2015], study 2 [Linsley et al., 2014], study 3 [Pabst et al., 2016]). (D) Low dimensionality representation (PCA based on the 1000 most variable genes) of the tumor- infiltrating cell gene expression profiles from different patients. Each point corresponds to cell-type average per patient of the single-cell RNA-Seq data of Tirosh et al. (2016) (requiring at least 3 cells of a given cell type per patient). Only samples from primary tumors and non-lymphoid tissue metastases were considered. Projection of the original single-cell RNA-Seq data can be found in Figure 1—figure supplement 1.] Immune cells differ in their gene expression profiles depending on their state and site of origin (e.g., blood or tumors) (Ganesan et al., 2017; Speiser et al., 2016; Zheng et al., 2017). To study the potential effect of these differences on our predictions, we established reference gene expression profiles of each major tumor-infiltrating immune cell type (i.e., CD4 T, CD8 T, B, NK, macrophages). We further derived reference profiles for stromal cells (i.e. cancer-associated fibroblasts (CAFs)) and endothelial cells. These reference gene expression profiles were obtained as cell type averages from