Chemotherapy induces myeloid-driven spatial T-cell exhaustion in ovarian cancer

Abstract

To uncover the intricate, chemotherapy-induced spatiotemporal remodeling of the tumor microenvironment, we conducted integrative spatial and molecular characterization of 97 high-grade serous ovarian cancer (HGSC) samples collected before and after chemotherapy. Using single-cell and spatial analyses, we identify increasingly versatile immune cell states, which form spatiotemporally dynamic microcommunities at the tumor-stroma interface. We demonstrate that chemotherapy triggers spatial redistribution and exhaustion of CD8+ T cells due to prolonged antigen presentation by macrophages, both within interconnected myeloid networks termed "Myelonets" and at the tumor stroma interface. Single-cell and spatial transcriptomics identifies prominent TIGIT-NECTIN2 ligand-receptor interactions induced by chemotherapy. Using a functional patient-derived immuno-oncology platform, we show that CD8+T-cell activity can be boosted by combining immune checkpoint blockade with chemotherapy. Our discovery of chemotherapy-induced myeloid-driven spatial T-cell exhaustion paves the way for novel immunotherapeutic strategies to unleash CD8+ T-cell-mediated anti-tumor immunity in HGSC.

Figure 1:. Multiomics characterization of the tumor-immune microenvironment of high-grade serous ovarian cancer

a Schematic overview of the data and workflow from clinicomolecular profiles, tumor-immune microenvironment architecture, single-cell states, spatial communities and programs to clinical correlations. b Dataset description and clinical features for 97 samples from 55 patients with HGSC. The samples which were SBS3 negative but BRCA mutated are marked with an asterisk. Patient age at diagnosis, tumor BRCA1/2 mutation or HRD status (Sig3, ovaHRDScar) were evenly distributed across the dataset and have been confirmed not to act as confounders of the results. c UMAP projection for cell types in t-CycIF data and d for scRNA-seq data. The cell type annotation colors for d are the same as for c. e UMAP of fine-grained cell types in scRNA-seq data. f Stacked bar plots of cell type proportions per sample for samples with patient- and site matched t-CycIF, scRNA-seq and RNA-seq data. Bar plots are ordered by increasing tumor cell percentage as determined from t-CycIF data. Bar plot annotations show cell type counts per t-CycIF and scRNA-seq cell types. g Heatmap showing the log2 fold change for the change in cell type proportions in IDS to chemo-naive samples in the different data modalities. Paired samples Wilcoxon test p-values with a value less than 0.05 for the comparison of IDS and chemo-naive samples per data modality are marked with an asterisk.

Figure 2:. Chemotherapy-associated dynamic changes in cell type proportions and ligand-receptor interactions

a Beeswarm plot showing the enrichment of cell state neighborhoods in chemo-naive and IDS samples. Light gray color denotes cell state neighborhoods with an FDR value > 0.05. b Circos plot depicting the top 50 ligand receptor interactions in chemo-naive samples and c in IDS samples d MultiNicheNet ligand-receptor interaction analysis showing the top 50 different ligand receptor interactions between macrophages and CD8+ T cells, their ligand activity, and scaled ligand activity e The change in expression of selected ligands and receptors from MultiNicheNet analysis in immune deconvoluted bulkRNAseq data from chemo-naive to IDS samples. In boxplots, the black horizontal lines represent the sample medians, the boxes extend from first to third quartile and the whiskers indicate the values at 1.5 times the interquartile range. Individual dots represent values per sample. f Spatial heterogeneity in cell-cell interactions depicted in t-cycif image and associated xy-plot with cell types colored, and contour plots of CD4+T-cell interactions with IBA1+CD11c+ macrophages as well as CD8+ T cell interactions with CD11c+ myeloid cells. g Heatmap with log2 fold change of expression values (color of the heatmap) of functional markers in macrophages interacting with CD8+T-cells as compared to without interacting CD8+T-cells in chemo-naive and IDS samples. Paired Wilcoxon test p-values <0.05 when comparing values of IDS and chemo-naive samples are marked with a red asterisk.

Figure 3.. Recurrent cellular neighborhoods uncover heterogeneous cellular architecture

a Stacked barplot of the cell type composition of each recurrent cellular neighborhood (RCN). The gray barplot on the right represents the total cell type count in each RCN. The final annotation on the right represents Rao’s entropy of each neighborhood. b Bar plot ordered by hierarchical clustering of the RCN proportions per sample clusters them into two groups. Annotations include the HRD status (red=HRP, blue=HRD) and treatment stage (blue=chemo-naive, yellow=IDS). c The heatmap of the logarithmic CD8+ T cell proportion per RCN separately in chemo-naive and IDS samples shows a more even CD8+ T cell infiltration pattern in IDS samples. Rows highlighted with black squares have a p-value < 0.05 in the FDR-corrected Wilcoxon paired test between chemo-naive and IDS samples. d Box plots showing higher closeness centrality score representing spatial dispersion for TIM3+CD8+T-cells as well as CD8+ T cells in IDS samples as compared to chemo-naive. In boxplots, the black horizontal lines represent the sample medians, the boxes extend from first to third quartile and the whiskers indicate the values at 1.5 times the interquartile range. Individual dots represent values per sample. e Representative image showing a t-CycIF image on the left, and corresponding RCN neighborhoods (large panel) and individual cell types (small boxes) on the right. f Representative t-CycIF image of myelonets on the left, and cells belonging to myelonets connected using Delaunay triangulation on the bottom (RCN14). g Heatmap with hierarchical clustering of the mean functional marker expression per individual myelonet with the patient of origin, HRD status, number of cells per network, and sample type as annotations h Dot plot and heatmaps showing differences in CD8+ T cell functional states across the RCN neighborhoods. The dot plot shows the log2 fold change in the mean marker expression in IDS to chemo-naive samples with FDR-corrected Wilcoxon p-values as the size of the dot, while heatmaps on the right show the column-wise z-score of marker expressions in CD8+ T cells in RCNs in chemo-naive and IDS samples, respectively.

Figure 4.. The tumor-stromal interface shows divergent CD8+T-cell and myeloid interaction patterns after chemotherapy

a A representative image of a tumor area and neighboring stroma with the corresponding RCN neighborhoods shown below the immunofluorescent image. The neighboring cells for subsequent interaction analyses were performed with a 45 micron radius as depicted by the white dashed circles. b Heatmap showing row-wise z-scores for cell type proportions belonging to RCN7 per sample clustered by hierarchical clustering. Annotations on the left include HRD status (red=HRP, blue=HRD) and treatment stage (blue=chemo-naive, yellow=IDS), the bar plot on the top shows the number of individual cell types belonging to the RCN7. c Dot plot showing changes in cell-cell interaction in chemo-naive and IDS samples in the RCN7. The y-axis represents the cell types and the x-axis their immune cell neighbors. The color of the dots represents the Wilcoxon test effect size, with values over 0 representing increased interaction probability in IDS samples, and values less than 1 in chemo-naive samples, respectively. The size of the dots represents the Wilcoxon p-value of the comparison. Those dots highlighted by black squares have an FDR<0.05 in the comparison between chemo-naive and IDS samples. d Representative images of the tumor-stromal interface and associated immune cell infiltrate in three chemo-naive and three IDS samples.

Figure 5.. GeoMx spatial transcriptomics reveals differences in pathway activities between regions of different CD8+ T-cells and IBA1+ myeloid cells abundance.

a Workflow b Left panel: A representative image of tissue slide stained for GeoMx spatial transcriptomics sequencing, with selected ROIs based on the relative abundance of CD8+T-cells and IBA1+ macrophages within a slide. Right panel: Representative images of each one of 4 ROI types. Top row displays a corresponding fragment of t-Cycif image of the adjacent tissue slide, used for guided ROI selection. The slide is stained with antibodies against Pan-CK, CD8 and IBA1. Middle row represents selected ROI on GeoMx tissue slide, stained with Sytox for cell nuclei and Pan-CK. Bottom row represents masks created over ROI, representing tumor and stromal areas (AOIs). c Clustered heatmaps representing mean z-score values of GSVA scores for 23 pathways. Columns represent each of 4 ROI types, rows are clustered based on values similarities between pathways. b represents stromal and tumor AOIs of chemo-naive and IDS samples c represents stromal and tumor AOIs of IDS samples from patients with long and short PFS. d Dot plots representing statistically significant differences (two-sided Wilcoxon rank-sum test) in GSVA scores between IDS vs chemo-naive samples for each of 4 ROI types and AOI compartments separately. Color represents the difference in mean GSVA score, while size of the dot represents the p-value of Wlcoxon test. Dot plots show the pathways, from Fig. 5 b,c which showed significant differences. e Violin plots representing differences in expression of selected genes within CD8+IBA1+ ROI type show differences between chemo-naive and IDS sample. Statistical significance was assessed using a two-sided Wilcoxon rank-sum test. Expression values are represented in log10 scale, with red dot denoting the mean value. f-i Circle plots showing significantly correlated (p-value <= 0.05, spearman correlation coefficient > 0.6) groups of pathways in CD8+IBA1+ AOIs from stromal and tumor compartments in chemo-naive and IDS samples separately. Each thin line represents one significantly correlated pathway from a specific pathway group (Supplementary table 4). Thick lines represent multiple pathways. Only pathways correlated with the T-cell exhaustion pathway group (general T-cell exhaustion, CTLA4 and PD1 signaling pathways) are shown. Each of the connections is shown twice - once in the color of the pathway group of interest (various colors) and once in the color of T-cell exhaustion group (light green), to highlight the bilateral nature of interactions.

Figure 6:. Immunocompetent spheroids show T-cell activation following treatment with anti-TIGIT, anti-PD1 and cisplatin

a General workflow illustrating the culturing process of patient-derived immuno-competent cultures (iPDCs) and subsequent flow cytometry analysis. b Representative immunofluorescence image of the cancer cell spheroids alongside infiltrating immune cells within the iPDCs. Blue: Hoechst, Red: CD45, Gray: CK7, Yellow: Death cell marker. Scale bar = 50 um c Barplots presenting the proportions of the various immune cell types within one iPDC. d Heatmap visualizing the Log2 Fold Change (FC) of Granzyme B expression in CD8+ T-cells of each tiragolumab-based immune checkpoint blockade (ICB) treatment condition compared to non-treated control. e Heatmap visualizing the Log2 Fold Change (FC) of Granzyme B expression in CD8+ T-cells of each tiragolumab-based ICB treatment condition combined with cisplatin compared to the corresponding ICB treatment condition alone. e Summary figure of findings regarding the differences of chemo-naive and IDS samples.