Temporal Clonal Tracing and Functional Perturbation Reveal Niche-Adaptive and Tumor-Intrinsic IFNγ Dependencies Driving Ovarian Cancer Metastasis

Abstract

Metastasis is an emergent continuum, driven by evolving reciprocal adaptations between continuously disseminating tumor cells (DTCs) and the specialized metastatic niches of distant organs. The interplay between intrinsic and niche-driven mechanisms that enables DTCs to survive and home to distant organs remains incompletely understood. Here, using MetTag, a single-cell barcoding and transcriptome profiling approach with time-stamped batch identifiers (BC.IDs), we mapped temporal, clonal dynamics of DTCs and the immune cell landscape across ovarian cancer metastatic niches. Deep sequencing of barcodes revealed preferred enrichment of early-disseminated clones across metastatic niches. Mechanistically, single-cell RNA sequencing (scRNA-seq) coupled with velocity analyses in ascites and metastasis-bearing omenta uncovered an emergent, distinct interferon-gamma (IFNγ) centric transcriptional trajectory, enriched among early seeding clones. Moreover, in vivo CRISPR/Cas9 screening of metastatic niche-specific signatures demonstrated that genes belonging to the ascites IFNγ signature, including Marco, Gbp2b, and Slfn1, are functionally important for peritoneal metastasis. Knockout of IFNγ receptor 1 (Ifngr1) in tumor cells significantly reduced metastatic burden and extended survival, underscoring the importance of tumor cell intrinsic IFNγ signaling in ovarian cancer metastasis. Furthermore, we identified that the tumor intrinsic IFNγ response and ascites-derived tumor-associated macrophages (TAMs) protect cancer cells from anoikis-mediated death within the IFNγ-rich ascites environment. Our study resolves temporal dynamics of disseminating tumor cells and highlights an ascites-driven, IFNγ program as a necessary pro-metastatic adaptation in the ovarian metastasis cascade.

Keywords: Ascites; Barcoding; CRISPR screening; Interferon; Lineage tracing; Metastasis; Omentum; Ovarian Cancer; Single cell analysis.

Figure 1:. Tracking Clonal Lineage in Ovarian Cancer Metastasis

(A) Schematic depicting main elements of the MetTag library barcoding construct. (B) Schematic of injection and tissue collection timeline for the staggered injection experimental model using MetTag-barcoded ID8 p53^−/− ovarian cell lines. Schematic was partially created with BioRender.com. (C) Genomic DNA (gDNA)-level barcode distributions in the omentum graphed as frequency under WT, NRG, and MAC depletion conditions. (D-E) gDNA-level barcode distributions in ascites and pelvic fat pads under WT and NRG conditions. In (C-E) panels, BC.ID1 = TP-1, BC.ID2 = TP-2, BC.ID3 = TP-3, BC.ID4 = TP-4, BC.ID5 = TP-5, BC.ID6 = TP-6 for clarity. (F) Injection schema for comparing one-time (400,000 total, n = 10) cell dose and fractionated (2,400,000 total, n = 10) cell injections (left) and representative BLI images at endpoint (right). (G) Bar graph depicting total flux emission in each group at experimental endpoint. (H) Schematic of scRNA-seq experiment conducted using a reversed BC.ID injection order. LARRY barcodes are readily captured by Chromium 10x Feature Barcoding. (I) Pie charts of RNA-level barcode frequencies in each location grouped as early (BC.ID5, BC.ID6), mid (BC.ID4, BC.ID3), and late (BC.ID2, BC.ID1). (J) Heatmap of sample level barcode frequencies, BC.ID6 = TP-1, BC.ID5 = TP-2, BC.ID4 = TP-3, BC.ID3 = TP-4, BC.ID2 = TP-5, BC.ID1 = TP-6. (K) Volcano plot of DEGs between early and late timepoint barcoded cancer cells where threshold for significance was defined at p <0.05 and log2FC < −1.0 for downregulation or >1.0 for upregulation. Genes labeled are the top 30 genes based on significance. Data in Figure 1G was analyzed using a two-tailed Student’s t test and graphed as mean with standard deviation (SD). In Figure 1K, the DEG table was generated using the Seurat FindMarkers() function and p values calculated via a Wilcoxon rank-sum test. For all Wilcoxon rank-sum tests, p values graphed represent the adjusted p value.

Figure 2:. Clonal Analysis Reveals Shared Evolutionary Path Towards Inflammatory Signature

(A) Split UMAP depicting cluster distributions across the four experimental conditions and identifying cancer cell-enriched clusters in ascites and omentum. (B) Reclustered cancer cell UMAP colored by the two metastasis-bearing anatomical locations. (C) Stacked bar chart depicting the frequency of Seurat subclusters within each metastatic location. (D) Seurat cluster UMAP of reclustered cancer cells with transcriptional program labels. (E) Dot plot depicting Hallmark pathways across all subclusters used to define each cluster subset gene program in (D). (F) Schematic representation of the possible outcomes of LARRY clonal analysis based on transcriptome patterns of enriched clones. (G) Heatmap of normalized clonal LARRY barcode counts across the two anatomical locations and individual mice (HTOs). (H) UMAP overlay of top 3 expanded clones in the omentum. (I) Bar chart depicting the relative subcluster proportions present in each of the top 3 barcode clones. (J) scVelo-generated PAGA velocity graph indicating trajectory arrows and inferred cluster lineage. (K) scVelo and CellRank-based simulated path from start to endpoint clusters. (L) Violin plot depicting RNA-level contribution of each timepoint to the IFNγ signature. Data in Figure 2L was analyzed by first defining the IFNγ signature using canonical mouse genes, followed by directly applying the wilcox.test() function in Seurat to compare TP-1 and TP-6 expression.

Figure 3:. Activation of IFNγ Response in Ascites is Necessary for Successful Omental Colonization

(A) DEG-based volcano plot of gene expression differences between OmMet and AscMet, where threshold for significance was defined at p <0.05 and log2FC < −1.0 for downregulation or >1.0 for upregulation. (B) Boxplots comparing IFN signature modules between AscMet and OmMet. (C) Schematic describing in vivo ovarian metastatic niche signature screen. Partially created with BioRender.com. (D) In vivo CRISPR screen RankView plots generated using MAGeCKFlute. OmMet (n=5) and AscMet (n=4) samples were compared to the original pool of ID8 p53^−/− cells. (E) Ranks of individual sgRNAs targeting positive control Pten and depleted genes. (F) Lollipop plot of depleted pathways associated with depleted genes in Fig. 3E. (G) IFNγ treatment time course qPCR plots of select CRISPR hits and positive controls. (H) IFNγ treatment time course western blot probing for Stat1/pStat1 and GBP1. (I) Schematic of in vivo experimental metastasis study comparing sgNTC and sgIFNGR1.g1 groups. (J) Time course log10-transformed total emission/flux values (left). Two independent groups pooled; cohort 1 NTC n = 8, KO n = 7; cohort 2 NTC n = 5, KO n = 5. log10-transformed flux plot for week six with p value (right). (K) Kaplan Meier survival curve generated for cohort 1. Data in Figure 3B was analyzed by first defining IFN signatures, followed by applying wilcox.test() in Seurat to compare AscMet and OmMet. CRISPR screen related data (3D-F) was first analyzed using MAGeCK. Count summary, gRNA summary, and gene summary output files were loaded into RStudio to generate rankview plots using MAGeCKFlute. For Figure 3G, qPCR data was analyzed by averaging technical replicates for each biological replicate, followed by calculating dCq by normalizing target gene Cq to housekeeping gene control. Next, 2^(-ddCq) or fold change was calculated by normalizing the values to the zero hour (PBS) treatment samples and plotted in Prism GraphPad. Significance between groups was assessed using a two-tailed Student’s t test and data was plotted as mean with standard deviation (SD). Data in figure 3J left panel was plotted as log transformed mean with 95% confidence interval (CI) and a two-tailed Student’s t test was performed at each timepoint. Data in 3J right panel was plotted with median log transformed values.

Figure 4:. IFN-γ Sensing Mediates Resistance to Anoikis and Activates Pro-survival Signaling

(A) Colony formation assay performed without (left) or with (right) 100 ng/ml IFNγ. (B-C) Cell-Titer Glo assay luminescence CTG values graphed as fold change for regular culture conditions (B) and ultra-low attachment conditions (C), with and without IFNγ treatment. Fold change was determined by normalizing each cell line’s CTG value to the regular attachment, no IFNγ treatment condition. (D) CTG flux value dot plots for Stat1-KO and Irf1-KO cells under ultra-low attachment, with or without IFN-γ treatment. (E) scRNA-seq based gating strategy used to identify Irf1-high and low cells for module scoring. (F) Violin plots for autophagy, stress, and pro-survival (anti-apoptotic) scores in Irf1-high vs. Irf1-low cells. Violin plots comparing Stat3 (G), Mcl-1, and Bcl2 (H) transcript levels between Irf1-high, mid, and low cells. (I) Western blot probing for Stat3 activation and SQSTM1 expression under ultra-low attachment conditions, with or without IFNγ treatments. (J) Violin plot of dormancy module comparing Irf1-high and low cells. CTG assay luminescence data (4B-4D) was analyzed in Prism using student’s t test. Data in B and C was graphed as mean and standard deviation, data in D was graphed as median values. Anoikis and dormancy modules, along with Stat3, Mcl-1, and Bcl2 transcript expression levels were compared between Irf1-high and Irf1-low cells by applying the wilcox.test() function in Seurat.

Figure 5:. Ascites and Omental TAMs Differ in Their Immunomodulatory Potential

(A) Violin plots depicting CRISPR screen hits expression level across Broad cell ID categories (cell types). (B) CellChat-generated circus plots of interaction strengths between Broad cell ID categories in metastatic omenta (left) and ascites (right). (C) Stacked bar chart of cell type proportions across the four experimental groups. (D) UMAPs of reclustered myeloid cells, split by experimental condition. (E) scVelo RNA velocity plot showing arrows pointing from ascites TAMs (clusters 1 and 5) to omentum TAMs (clusters 0 and 6). (F) DEG volcano plot comparing gene expression differences between ascites TAM cluster 1 and omentum TAM cluster 0. (G) Stacked bar chart of subsetted T/NK cells comparing cell proportions in ascites and omentum. (H-I) ORA generated Reactome pathways enriched in ascites (left) and omentum-derived TAMs (right). Significance of DEGs in Figure 5F was evaluated by applying the FindMarkers() and Wilcoxon test in Seurat to compare ascites and omental Cd68^high/Adgre1^high macrophages.

Figure 6:. NicheTracing Identifies Ascites-derived TAMs as Pro-Metastatic Interactors

(A) Schematic describing the Cre-based NicheTracing labeling system where ID8 p53^−/− and E0771 cancer cells express Cre.ERT2 which transfers to neighboring RCL-ZsGreen niche cells. Upon nuclear entry of Cre.ERT2, labeled niche cells turn ZsGreen positive (ZsG+). (B) Representative gating strategy depicting the major cell population tagged with zsG. (C-D) Bar plots showing preference for F4/80+ macrophage labeling in ascites. (E) Liposomal chlodronate-based macrophage depletion strategy and timeline schematic with representative tissue images. n= 7 liposome control group and n = 7 chlodrosome treatment group. (F) Bar plot showing quantification of omental metastases via omental weight measurement at endpoint. (G) Images of ascites samples harvested at endpoint for flow cytometry analysis (left) and flow-based quantification of CD45 negative cells (right). (H) Schematic of macrophage-cancer cell coculture layout (top) and representative coculture images for one cell line (bottom). (I-K) Plots of percent CD45 negative cancer cell viability post coculture. Data in Figures 6C–D and 6F–6H was plotted as mean with SD and differences between groups were compared via a two-tailed unpaired Student’s t test. Data in Figures 6I–K was similarly analyzed using a Student’s t test and plotted with median values.

Figure 7:. Patient Ascites MAC/Tumor Interactions are Enriched for Tumor-supporting Pathways

(A) Stacked bar chart of main cell type proportions captured in human scRNA-seq dataset across five anatomical locations, using maintype-2 annotation provided by publisher. (B) Heatmap depicting DEGs between Met.Ome and Ascites. (C-D) GO Molecular Function pathway level differences between omental MACs (C) and ascites MACs (D). (E) Schematic depicting NicheNetR analysis approach for finding top TAM ligands in both murine and human datasets. This panel was partially generated using BioRender.com. (F) Venn diagram of top NicheNet predicted ligands after filtering for robust expression and overlapping with top interactions in our murine dataset. (G) Heatmap showing top receptor-ligand interactions upregulated in patient ascites-MACs compared to metastases. (H) GO enrichment dotplot generated from the gene list containing top target genes in cancer cells that are predicted to respond to top ligands in the human dataset. Data in Figure 7B was analyzed using the FindMarkers() function in Seurat and Wilcoxon test to compare Asc and Met.Ome.