Gene expression markers of Tumor Infiltrating Leukocytes

Abstract

Background: Assays of the abundance of immune cell populations in the tumor microenvironment promise to inform immune oncology research and the choice of immunotherapy for individual patients. We propose to measure the intratumoral abundance of various immune cell populations with gene expression. In contrast to IHC and flow cytometry, gene expression assays yield high information content from a clinically practical workflow. Previous studies of gene expression in purified immune cells have reported hundreds of genes showing enrichment in a single cell type, but the utility of these genes in tumor samples is unknown. We use co-expression patterns in large tumor gene expression datasets to evaluate previously reported candidate cell type marker genes lists, eliminate numerous false positives and identify a subset of high confidence marker genes.

Methods: Using a novel statistical tool, we use co-expression patterns in 9986 samples from The Cancer Genome Atlas (TCGA) to evaluate previously reported cell type marker genes. We compare immune cell scores derived from these genes to measurements from flow cytometry and immunohistochemistry. We characterize the reproducibility of our cell scores in replicate runs of RNA extracted from FFPE tumor tissue.

Results: We identify a list of 60 marker genes whose expression levels measure 14 immune cell populations. Cell type scores calculated from these genes are concordant with flow cytometry and IHC readings, show high reproducibility in replicate RNA samples from FFPE tissue and enable detailed analyses of the anti-tumor immune response in TCGA. In an immunotherapy dataset, they separate responders and non-responders early on therapy and provide an intricate picture of the effects of checkpoint inhibition. Most genes previously reported to be enriched in a single cell type have co-expression patterns inconsistent with cell type specificity.

Conclusions: Due to their concise gene set, computational simplicity and utility in tumor samples, these cell type gene signatures may be useful in future discovery research and clinical trials to understand how tumors and therapeutic intervention shape the immune response.

Keywords: Cell types; Gene expression; Immunotherapies; TILs; Tumor infiltrating lymphocytes.

Fig. 1

Simulated example of process for evaluating candidate genes for marker-like co-expression in a tumor dataset. a Example of marker-like co-expression. Genes 1 and 2 are highly correlated with a slope of 1, a pattern consistent with both genes rising and falling at the same rate. b Example of two genes which cannot both be markers. Gene 3 exhibits no co-expression with gene 1, showing that Genes 1 and 3 are not both good markers for the same cell type. c Example of two genes which cannot both be markers. Gene 4 is highly correlated with Gene 1, but with a slope different than 1, meaning they are not both markers for the same cell type

Fig. 2

Pairwise similarity, a measure of marker-like co-expression, of candidate B-cell marker genes in TCGA. a Pairwise similarity of candidate B-cell marker genes averaged across 24 TCGA RNASeq datasets. Darker red indicates co-expression patterns consistent with both genes acting as cell type markers. Values of 1 indicate perfect marker-like co-expression. Green sidebars indicate final selected markers. b Two of the selected B-cell markers, including CD19, in the bladder cancer dataset, demonstrating strong marker-like co-expression. c In bladder cancer, CD19 and the rejected candidate marker BLNK, displaying co-expression inconsistent with both genes acting as B-cell markers

Fig. 3

Pairwise similarity, a measure of marker-like co-expression, of selected marker genes in TCGA. Values of 1 indicate perfect marker-like co-expression. a Mean log2 expression vs. average pairwise similarity of selected cell type markers across TCGA datasets. Cell types in grey have been discarded from the final panel of markers. b Average pairwise similarity of each cell type’s marker genes in each TCGA dataset

Fig. 4

Comparison of gene expression cell scores to alternative biomarkers. a In FFPE tumor samples, gene expression cell scores and log2-transformed IHC measurements of cell type abundance. b In PBMC samples, gene expression and flow cytometry measurements, normalized to T cell abundance

Fig. 5

Reproducibility of cell scores derived from triplicate runs of 12 tumor samples. For each cell type, each sample’s individual replicate cell scores are plotted against its average score. Color denotes tumor type

Fig. 6

Results of analyses of cell scores in TCGA RNASeq data. a Boxplot of Total TILs score across TCGA datasets. Datasets are ordered according to median score. The vertical axis is log2-scale. b Prognostic information in Total TILs score and in cell type enrichment scores, the residuals of each cell score when regressed on the Total TILs score. Red indicates cell types whose enrichment within the total infiltrate is associated with poor outcome; blue indicates association with good prognosis. Only results with FDR < 0.05 are shown, and cancers without any statistically significant cell types are not shown

Fig. 7

Application of cell scores to an immunotherapy dataset. a Total TILs score of each biopsy at each timepoint. Total TILs score was calculated as the average of all cell scores with >0.6 correlations with CD45, a metric that excluded only NK cells and neutrophils. Grey points denote non-responders; colored points denote responders. b Estimates and 95% confidence intervals for each cell score’s log2 fold-change between responders and non-responders on anti-CTLA4 (top) and anti-PD1 (bottom). c Cell scores from a single anti-CTLA4 responder before and during therapy. Scores are given as log2 fold changes from the average patient’s pre-treatment score