Varadha Balaji Venkadakrishnan

Table S8 from Vitronectin Enrichment in Prostate Cancer Liver Metastases Promotes Adhesion and Survival

Table S7 from Vitronectin Enrichment in Prostate Cancer Liver Metastases Promotes Adhesion and Survival

Table S6 from Vitronectin Enrichment in Prostate Cancer Liver Metastases Promotes Adhesion and Survival

The development of liver metastases in prostate cancer is associated with aggressive disease and poor prognosis. Because hepatocytes exhibit high metabolic activity with unique secretory profiles, we hypothesized that hepatocyte-to-tumor cell signaling plays a role in promoting liver metastasis. We evaluated single-cell transcriptomic data and metastatic tissue from patients with castration-resistant prostate cancer spanning androgen receptor (AR)–positive and AR-negative pathologies. Despite extensive intra- and intersample heterogeneity, communication analysis predicted vitronectin (VTN) engagement of tumor integrins as a common feature. VTN-positive hepatocytes were observed in prostate tumors with VTN accumulation in sinusoidal patches. Consistent with integrin activation, VTN treatment of AR-positive and AR-negative prostate cancer cells significantly promoted tumor cell adhesion, inhibited hypodiploid accumulation in accordance with survival effects, and stimulated FAK-dependent phosphorylation of ERK and AKT. FAK inhibition with defactinib abrogated VTN- and serum-mediated adhesion in a cell-specific manner and mitigated VTN-driven depletion of hypodiploid populations. These findings support a model in which dense intrasinusoidal VTN deposits capture metastatic prostate tumor cells in the liver and biochemically activate tumorigenic signaling, promoting tumor aggressiveness. Significance:Through the synthesis of single-cell transcriptomic data with experimental models, we identify VTN as a candidate microenvironmental mediator of prostate cancer liver colonization. VTN drives cell adhesion, inhibits hypodiploid accumulation consistent with cell survival, and activates oncogenic signaling, which may be inhibited by the dual FAK/PYK2 inhibitor defactinib.

Liver metastases are highly heterogeneous with implications for tumorigenicity and cell signaling. A, Left, Harmony-integrated liver metastatic samples (n = 13,488 cells) from GSE210358 colored by Seurat cluster (resolution = 0.15). Each sample is derived from a unique patient; Right, pie and bar charts depicting the number of cells in each cluster and the fraction of cells belonging to each patient in each cluster, respectively. B, Cell-type cluster annotations determined by manual annotation, gene list enrichment analysis of top upregulated genes, and CopyKAT copy-number inference. C, Top marker genes for each cluster. Significance was determined with the FindAllMarkers function configured to apply a Bonferroni-corrected Wilcoxon rank-sum test. Significantly upregulated genes were ordered by adjusted P value, and the first four genes were selected for display. D, The parent Seurat object was subset to tumor cells (n = 6,226 cells), and differential expression analysis was conducted on every tumor cluster using the FindAllMarkers function. Resulting markers were subset to genes of interest and plotted. FindAllMarkers was configured to use a Bonferroni-corrected Wilcoxon rank-sum test for statistical determinations. E, Average log FCs of tumor marker genes were inputted to the fgsea function for GSEA, which uses a false discovery rate–adjusted permutation test to calculate significance. Cytosk., cytoskeleton; Metab., metabolism; NES, normalized enrichment score; ns, not statistically significant.

Tumor-directed VTN signaling is a shared across clusters. A, CellChat results were subset to communication predicted to originate from hepatocyte-annotated cluster 7 and be received by tumor-annotated clusters 0, 3, and 4. Mean CellChat probabilities were calculated for each pathway across tumor clusters, pathways were ordered by their mean value, and the top 60 were selected for plotting. B, Ligands for the top 20 pathways in (A) were retrieved from CellChat results and annotated with tissue-level expression from The Human Protein Atlas bulk RNA-seq data. Tissue specificity was calculated for each ligand with the Tau index, and pathways were ordered by the mean Tau index of their ligands. C, CellChat-predicted ligand–receptor pairs involved in VTN signaling were retrieved from the CellChat object. CellChat probability scores and P values (permutation test) were retrieved for each pair and communicating cluster and subset to significant (P < 0.05) hepatocyte-to-tumor communication. Ligand–receptor pairs were ordered by their median CellChat probability score across tumor clusters for plotting. D, Single-cell normalized expression data were extracted for CellChat-annotated ligand–receptor pairs involved in VTN signaling. Boxplots show the distribution of cell-level expression for hepatocytes and tumor cells, and points show cluster means. Significance was calculated with DESeq2 from sample and tumor/hepatocyte-level pseudobulked counts. ns, not statistically significant. E, UMAPs colored by binary VTN, ITGB1, or ITGAV expression. F and G, Count and FPKM matrices were obtained from GSE211448 and cBioPortal for the 22Rv1 orthotopic model and SU2C Dream Team datasets, respectively. For 22Rv1, DESeq2 was applied to normalize counts and calculate FCs between 22Rv1-LMD and primary 22Rv1. For SU2C Dream Team data, FCs were calculated from pseudocount-adjusted FPKMs. Then, FCs were inputted to fgsea for GSEA, which uses a false discovery rate–adjusted permutation test to calculate significance. Enrichment results for Reactome integrin signaling (R-HSA-354192) are shown. H, Liver tissue was harvested at autopsy (Supplementary Table S8) and formalin-fixed, paraffin-embedded, and sectioned. Slides were stained with hematoxylin and eosin (H&E), relevant lineage markers, and VTN. ***, P < 0.001.

Tumor subtypes present distinct communication patterns. A, CellChat was used to infer intercellular communication from cluster ligand, receptor, and cofactor expression. CellChat uses a permutation test for statistical determinations, and all results depict significantly enriched pathways (P < 0.05). Plotted pathways were selected by averaging pathway-level CellChat probability across all groups and selecting pathways with means in the top 50th percentile. B, CellChat probability scores were averaged for each significantly enriched pathway across communication direction. For each direction, the top 30 pathways were selected. Due to overlap between pathways, 44 unique pathways are shown. C, CellChat probability scores were averaged or tallied for each communicating pair across significantly enriched pathways. All 10 predicted communicating pairs are shown. D, CellChat results were subset to communication between each pair (regardless of communication direction), and CellChat probability scores were averaged across communication direction for each interaction; scores for monodirectional interactions (i.e., autocrine signaling) have no associated standard error (SE) in this context. Interactions were ordered by mean probability, and the top 10 were selected for plotting.

The adhesion effects of VTN are robust across assay conditions and biological sources. A, Representative microscopy of prostate cancer cells plated on polyHEMA-coated plates depicting the abrogation of cell attachment and successful spheroid formation. B, Adhesion assays were performed with polyHEMA-coated plates and serum-free media with or without 5 μg/mL VTN. Bars show three biological replicates, and significance was calculated with two-tailed Welch t tests. ns, not statistically significant. C, Cells were subjected to a time-course adhesion assay in serum-free media with or without 5 μg/mL VTN. Bars show three biological replicates, and significance was calculated for the 24-hour timepoint with two-tailed Welch t tests. D, Images were taken of well centers from the 24-hour timepoint, and cells within captured fields of view were counted. Results are for three biological replicates, and two-tailed Welch t tests were used for statistical determinations. E, Adhesion assays were performed in serum-free media with 5 μg/mL rVTN, 5 μg/mL human plasma–purified VTN (hVTN), or no VTN. Complete, use of complete media as a positive control. Bars show three biological replicates, and significance was calculated with two-tailed Welch t tests. F, Conditioned media was recovered from 5-donor human plateable hepatocytes (first collection period/day 1). An adhesion assay was performed with 22Rv1 and 100% conditioned media serially diluted 1:1 for seven timepoints. Images show representative regions of adhered or suspended 22Rv1 cells following exposure to 100%, 50%, 25%, or 0% conditioned media. G, Conditioned media (24 μL) was directly mixed with 8 μL of 4× SDS-PAGE and immunoblotted for VTN as specified in “Materials and Methods.” As media was collected and replenished every 24 hours, lane labels indicate the period of media collection since hepatocyte plating (day 1 corresponds to collection on day 1 since plating). Uncond. media, unconditioned hepatocyte maintenance media. H, An adhesion assay was performed with basal hepatocyte maintenance media, 50% hepatocyte-conditioned maintenance media, or complete culture media as a positive control (RPMI-1640 with 10% FBS). Complete, use of complete culture media as a positive control; maintenance, use of maintenance media as the basal serum-free media (rather than RPMI-1640). Bars show three biological replicates, and significance was calculated with two-tailed Welch t tests. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001.

VTN localizes to the hepatic sinusoid lumen and promotes prostate cancer adhesion and spreading. A, Representative sinusoidal or cytoplasmic VTN staining from autopsy-harvested tumor tissue. B, Rapid autopsy tissue from DFCI-HB-264 was immunostained for CD31 (red) and VTN (green). Fluorescent images were loaded into QuPath, and CD31+ regions were defined with the native pixel classification thresholder algorithm. Individual CD31/red and VTN/green channels are shown, as well as a merged image in which CD31+ regions are outlined in yellow. C, From IHC staining, 20 representative cytoplasmic and sinusoidal regions per patient were marked with points. Point-level 3,3′-diaminobenzidine (DAB) intensity was averaged within each patient and staining pattern. Bars depict four biological replicates (from the four patients), and statistical indications show the results of a paired two-tailed t test. D, From immunofluorescence staining of DFCI-HB-264 liver tissue, within CD31+ regions and across the whole image, VTN+ areas were defined with the QuPath pixel classification thresholder algorithm retaining parent annotations (if a VTN+ region is within a CD31+ region). VTN intensities were normalized to the maximum value, regions were filtered to those with areas greater than 30 μm2, and values were compared between CD31+ (sinusoidal) and CD31− (cytoplasmic) regions. Each point corresponds to one within-section region (n = 23 cytoplasmic; n = 32 sinusoidal), and statistical determinations were made with a two-tailed Welch t test. E, Schematic outlining an assay for cell adhesion with soluble or adherent independent variables. F, The described adhesion assay was performed on cells in serum-free media with or without 5 μg/mL VTN and allowed to adhere for 24 hours (48 hours for VCaP). For all cell lines except PM154 and PM155: Complete, use of complete media as a positive control; for PM154 and PM155, complete, plating on collagen-coated plates. Bars show three biological replicates, and significance was calculated with two-tailed Welch t tests. ns, not statistically significant; OE, overexpression; WT, wild-type. G, Representative images depicting VTN-associated morphologic transitions after plating in serum-free media. Liv. Met., liver metastatic 22Rv1 from orthotopic or PROX1-expressing models; Pri., 22Rv1 PROX1-expressing cells. H, Cells of epithelial and spherical morphology from representative images of attached VTN or control-treated cells after plating with serum-free media were counted and compared. Statistical determinations were made from individual images taken from the same experiment with a two-tailed Fisher exact test. [E, Created in BioRender. Egelberg, J. (2026) https://BioRender.com/5blud6q.] *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.

VTN promotes cell survival and stimulates MAPK and PI3K. A, Cells undergoing active culture were aliquoted, pelleted, and lysed, and immunoblotting was performed as specified in “Materials and Methods.” ITG, integrin. B, Cells were plated in 6-well plates in 2 mL serum-free media with either 5 μg/mL VTN in dH2O (for C4-2, 22Rv1, PC-3, and DU145), 16.8 μg/mL VTN in dH2O (for PM154 and PM155, necessary to facilitate spreading), or dH2O alone and allowed to adhere for 24 hours. Cells were harvested for PI staining followed by flow cytometry, and data were gated to exclude debris and doublets. Hypodiploid cell fractions were calculated as described in “Materials and Methods,” and statistical significance was determined from individual representative experiments with a two-tailed Fisher exact test. ns, not statistically significant. C–H, Cells were plated on 6-well plates, serum-starved overnight, and exposed to VTN for the specified amounts of time. PM154 and PM155 were stimulated with the same concentration of VTN as the other cell lines. After exposure, cells were lysed, and immunoblotting was performed as specified in “Materials and Methods.” Red triangles indicate the pERK bands. ***, P < 0.001.