An analysis of ovarian cancer found that tumors with stronger SPP1-related immune signals were associated with shorter overall survival, while laboratory experiments suggested that lowering PI3 reduced tumor-cell SPP1 and was linked to a more inflammatory pattern in cocultured macrophages. The findings do not show that SPP1 causes poorer outcomes or that PI3 is a clinical treatment target.
The work combined single-cell data from 12 ovarian tissue samples, including seven treatment-naïve high-grade serous ovarian cancer tumors and five age-matched nonmalignant tissues, with TCGA-OV data and three independent validation cohorts. It also paired ES-2 and SKOV3 cancer-cell tests with macrophage cocultures and examined stable PI3 knockdown in an ES-2 xenograft model.
The tumor ecosystem looked different
After quality control, 54,701 of 65,820 cell barcodes remained. The analysis grouped them into 14 clusters and eight major cell types. Compared with nonmalignant tissue, high-grade serous ovarian cancer samples contained higher proportions of epithelial and macrophage cells and lower proportions of natural killer, endothelial and fibroblast cells. T-cell and B-cell proportions did not differ significantly.
Within the macrophage population, the researchers defined four states. The inflammatory group was less abundant in tumors, while the activated group was more common. The reported comparison P values were 0.006 and 0.009; differences for resident macrophages and SPP1-positive tumor-associated macrophages were not statistically significant, with P = 0.104 for both.
CellChat, a computational method that infers possible cell-to-cell signals from ligand and receptor transcripts, identified SPP1 interactions with CD44 and integrins across macrophage subtypes. In tumors, the inferred network involved tumor, endothelial, fibroblast and macrophage populations, with SPP1-positive tumor-associated macrophages remaining a central node. Because it is based on transcript-level inference, the result does not by itself demonstrate physical binding, receptor activation or spatial contact.
The signal tracked with survival
Using a fixed 11-gene SPP1-related set, the analysis split cases into two stable groups: SPP1-Low and SPP1-High. The high group had a higher pathway score and shorter overall survival. The estimated hazard of death, a measure of how quickly deaths accumulated over follow-up, was 1.74 times higher in SPP1-High than in SPP1-Low, with a 95% confidence interval from 1.35 to 2.23 and P < 0.001. The pathway-score difference had P = 0.0154.
The two molecular groups also carried different inferred immune profiles. SPP1-High showed stronger immune-related activity and higher estimated levels of memory B cells, M2 macrophages and resting dendritic cells. SPP1-Low was more associated with cell-cycle, mismatch-repair, ciliary-movement and metabolic pathways. These immune-cell figures are computational estimates rather than direct cell counts.
The researchers then compared four survival algorithms to build a risk score from TCGA-OV: LASSO-Cox, GBM-Cox, random survival forest and ABESS-Cox. GBM-Cox performed best by concordance index, a measure of how often a model correctly orders patients by survival time, at 0.768, versus 0.660, 0.652 and 0.630 for the other methods. High-risk groups had shorter survival in TCGA-OV and three validation datasets, with hazard ratios of 1.85 in TCGA-OV, 1.94 in GSE140082, 1.54 in GSE26712 and 1.41 in GSE13876.
PI3 effects stayed inside the lab
In cell tests, silencing PI3 in ES-2 and SKOV3 cells reduced SPP1 messenger RNA and its secretion. The change was accompanied by lower proliferation, wound closure, migration and invasion. In the ES-2 xenograft model, stable PI3 knockdown was also associated with slower tumor growth and lower terminal tumor volume and weight.
Macrophages cocultured with the PI3-silenced cancer cells showed higher TNF, IL1B, CXCL10 and CD86 expression and lower CD163, MRC1, CCL18 and IL10 expression. TNF-alpha and IL-12p70 increased, while IL-10 and CCL18 decreased, a pattern consistent with an inflammatory or M1-like state.
That makes the findings a starting point rather than a clinical answer. The SPP1 network was inferred mainly from transcript-level ligand and receptor data, so it did not directly demonstrate protein binding, receptor activation or spatial proximity. The macrophage experiments used coculture models rather than patient-derived macrophages, and the prognostic model requires prospective validation with more complete clinical information.
Paper data and sources
Original title: An SPP1-centered immune–parenchymal communication axis links macrophage heterogeneity to ovarian cancer prognosis
Authors: Liu X, Xia L, Sheng J et al.
Journal/Repository: Frontiers in immunology
Status: Peer-reviewed
First online: 2026-10-07
DOI: Not available
Original paper · Full text