Immune cell landscapes are associated with high-grade serous ovarian cancer survival

Abstract

High-grade serous ovarian cancer (HGSOC) is an aggressive disease known to develop resistance to chemotherapy. We investigated the prognostic significance of tumor cell states and potential mechanisms underlying chemotherapy resistance in HGSOC. Transcriptome deconvolution was performed to address cellular heterogeneity. Kaplan-Meier survival curves were plotted to illustrate the outcomes of patients with varying cellular abundances. The association between gene expression and chemotherapy response was tested. After adjusting for surgery status and grading, several cell states exhibited a significant correlation with patient survival. Cell states can organize into carcinoma ecotypes (CE). CE9 and CE10 were proinflammatory, characterized by higher immunoreactivity, and were associated with favorable survival outcomes. Ratios of cell states and ecotypes had better prognostic abilities than a single cell state or ecotype. A total of 1265 differentially expressed genes were identified between samples with high and low levels of C9 or CE10. These genes were partitioned into three co-expressed modules, which were associated with tumor cells and immune cells. Pogz was identified to be linked with immune cell genes and the chemotherapy response of paclitaxel. Collectively, the survival of HGSOC patients is correlated with specific cell states and ecotypes.

Keywords: Cell ratios; High-grade serous ovarian cancer; Macrophage; Overall survival; T cell activation.

Figure 1

Schematic workflow of our study. The high-grade serous ovarian cancer (HGSOC) dataset was downloaded from the NCBI GEO database. The expression matrix was transformed into a cell states matrix and ecosystems matrix by Ecotyper. Then, the differential cell abundance between dead and alive patients was identified. Survival analysis based on cell states or carcinoma ecotypes was performed. Differentially expressed genes (DEGs) between two patients split by the survival time were identified. Weighted gene co-expression analysis (WGCNA) based on DEGs was executed. WGCNA modules were annotated by functional enrichment analysis. Ratios of cell states and carcinoma ecotypes were calculated for ratio-based survival analysis. Finally, potential mechanisms of therapeutic benefit with paclitaxel were inferred.

Figure 2

Heatmap depicting 5, 8, 6, 6, 9, and 3 cell states identified from digitally purified (A) natural killer cells (NKs), (B) dendritic cells (DCs), (C) plasma cells (PCs), (D) epithelial cells (ECs), (E) monocytes/macrophages (MCs), and (F) CD8 T (CD8T) cells, respectively. For each cell type, the left panel represents the cell states of the TCGA dataset, while the right panel represents the cell states of the current HGSOC dataset. Patient samples (columns) are organized by the most prevalent cell state and annotated with bulk tumor cell-of-origin labels.

Figure 3

Overall survival (OS) for cell states (A) natural killer (NK) cells S4, (B) dendritic cells (DCs) S3, (C) plasma cells (PCs) S1, (D) epithelial cells (ECs) S4, (E) NK cells S1, and (F) monocytes/macrophages (MCs) S3 in the 260 HGSOC. Significance was determined by a two-sided log-rank test. The unit of time is the month.

Figure 4

Ecotypes in HGSOC. (A) Cell-state abundance profiles are organized into ten carcinoma ecotypes (CEs). Only cell states and HGSOC samples assigned to CEs are shown. HGSOC samples are ordered by the most abundant CE class per specimen. (B) CE9 and (C) CE10 are associated with patient survival in the HGSOC dataset GSE32062. The unit of time is the month.

Figure 5

The CE9- or CE10-high expression samples have distinct expression patterns. (A) A total of 1265 differential expression genes (DEGs) are identified. (B) The clustering heat map based on DEGs shows the distinct expression patterns between groups. In the color bar above the heat map, yellow indicates low CE9- or CE10-low expression samples, while red indicates CE9- or CE10-high expression samples. (C) When power was set at 12, the constructed co-expression network based on DEGs followed a power law. (D) The cluster dendrogram shows the assignment of genes into gene modules with different colors. The color bar indicates the assignment of genes to a module. (E) GO biological process enrichment and (F) KEGG enrichment analysis of module M2 genes by clusterProfiler.

Figure 6

Cell state ratios and ecotypes ratios in HGSOC. Cell state ratios of (A) fibroblasts S3/polymorphonuclear leukocytes (PMNs) S3, (B) CD8 T cell S3/mast cell S5, and (C) CD8 T cell S3/NK cell S3 are associated with HGSOC survival. Ecotype ratios of (D) CE1/CE9, (E) CE6/CE9, and (F) CE6/CE10 are associated with HGSOC survival. The unit of time is the month.

Figure 7

Genome-wide mutation analysis in the TCGA ovarian cancer dataset revealed the association between three module M2 hub genes (A) Sash3, (B) Traf3ip3, and (C) Ptprc and candidate regulatory gene Pogz. (D) Pogz expression is associated with patient overall survival. (E) The expression of Pogz across 22 different tissues of tumor and normal samples. (F) Pogz low expression patients response to paclitaxel treatment. (G) Pogz could separate samples from non-responders and responders to paclitaxel treatment.